cv::connectedComponentsWithStats使用4连通出错分析
一、问题描述
当我在C++多线程环境下使用OpenCV的cv::connectedComponentsWithStats函数,参数使用4连通(使用8连通没问题),但是报出了如下的错误:
Program terminated with signal SIGSEGV, Segmentation fault. #0 0x00007f2a3a2a1a67 in cv::connectedcomponents::LabelingBolelli4CParallel<int, unsigned char, cv::connectedcomponents::CCStatsOp>::SecondScan::operator()(cv::Range const&) const () from /home/mulan/MulanAlgo/deploy_env/opencv/lib/libopencv_imgproc.so.405
二、问题分析
2.1 连通组件标记算法介绍
连接组件标记算法(connected component labeling algorithm)是图像分析中最常用的算法之一,算法的实质是扫描一幅图像的每个像素,对于像素值相同的分为相同的组(group),最终得到图像中所有的像素连通组件。扫描的方式可以是从上到下,从左到右,对于一幅有N个像素的图像来说,最大连通组件个数为N/2。扫描是基于每个像素单位,对于二值图像而言,连通组件集合可以是V={1|白色}或者V={0|黑色}, 取决于前景色与背景色的不同。对于灰度图像来说,连图组件像素集合可能是一系列在0 ~ 255之间k的灰度值。
连通域分析一般应用于对区域分割后的处理。在需要将前景目标提取出来以便后续进行处理的应用场景中都能够用到连通区域分析方法,通常连通区域分析处理的对象是一张二值化后的图像。
OpenCV中支持连通组件扫描的API有两个,一个是带统计信息一个不带统计信息。
不带统计信息的API及其解释:
int connectedComponents( InputArray image, // 输入二值图像,黑色背景 OutputArray labels, // 输出的标记图像,背景index=0 int connectivity = 8, // 连通域,默认是8连通 int ltype = CV_32S // 输出的labels类型,默认是CV_32S )
该函数对图像的每个区域分析,然后将背景标记为0,其他的区域用从1开始的正整数依次标记。最后将标记结果返回给labels。
带统计信息的API及其解释如下:
int connectedComponentsWithStats( InputArray image, // 输入二值图像,黑色背景 OutputArray labels, // 输出的标记图像,背景index=0 OutputArray stats, // 统计信息,包括每个组件的位置、宽、高与面积 OutputArray centroids, // 每个组件的中心位置坐标cx, cy int connectivity, // 寻找连通组件算法的连通域,默认是8连通 int ltype, // 输出的labels的Mat类型CV_32S int ccltype // 连通组件算法 )
->其中stats包括以下枚举类型数据信息:
- 区域的外接矩形左上角点像素点坐标的X位置:CC_STAT_LEFT
- 区域的外接矩形左上角点像素点坐标的Y位置:CC_STAT_TOP
- 区域外接矩形的宽度:CC_STAT_WIDTH
- 区域外接矩形的高度:CC_STAT_HEIGHT
- 区域的面积(像素单位)CC_STAT_AREA
该函数对图像的每个区域分析,除了对区域和背景添加标记外(结果返回给labels),还可以获取每个区域的统计信息:区域的数目,外接矩形大小,面积,中心位置。(结果返回给stats)
->连通组件算法的连通域,可以是4连通和8连通:
- 四连通: 包括了中心点的上下左右四个方向的点
- 八连通:包括周围8个点的关系
2.2 错误分析
参考1:New Bing给出的回答
参考2:chatGPT给出的回答
为什么cv::connectedComponentsWithStats在多线程场景中使用4-connection可能会导致分割错误,但8-connection可以正常工作,这可能是由于OpenCV中函数的内部实现。 在OpenCV中,4连通组件标记算法使用两步算法对图像进行水平扫描,然后垂直扫描,而8连通组件标记算法使用单步算法对图像进行对角线扫描。 8连通算法中的对角扫描可以帮助避免某些边缘情况,提高整体标注的准确性。 在多线程场景中,图像被多个线程分割和处理的方式也会影响函数的性能和稳定性。 当多个线程同时访问共享数据时,4连接算法中图像的划分方式可能会导致问题,而8连接算法可能更能适应这些问题。
那么我们来看看源代码:
但是它给出的源代码与官网的还是有差别的,我们还是直接看正版:
/*M/////////////////////////////////////////////////////////////////////////////////////// // // IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING. // // By downloading, copying, installing or using the software you agree to this license. // If you do not agree to this license, do not download, install, // copy or use the software. // // // Intel License Agreement // For Open Source Computer Vision Library // // Copyright (C) 2000, Intel Corporation, all rights reserved. // Third party copyrights are property of their respective owners. // // Redistribution and use in source and binary forms, with or without modification, // are permitted provided that the following conditions are met: // // * Redistribution's of source code must retain the above copyright notice, // this list of conditions and the following disclaimer. // // * Redistribution's in binary form must reproduce the above copyright notice, // this list of conditions and the following disclaimer in the documentation // and/or other materials provided with the distribution. // // * The name of Intel Corporation may not be used to endorse or promote products // derived from this software without specific prior written permission. // // This software is provided by the copyright holders and contributors "as is" and // any express or implied warranties, including, but not limited to, the implied // warranties of merchantability and fitness for a particular purpose are disclaimed. // In no event shall the Intel Corporation or contributors be liable for any direct, // indirect, incidental, special, exemplary, or consequential damages // (including, but not limited to, procurement of substitute goods or services; // loss of use, data, or profits; or business interruption) however caused // and on any theory of liability, whether in contract, strict liability, // or tort (including negligence or otherwise) arising in any way out of // the use of this software, even if advised of the possibility of such damage. // // 2011 Jason Newton <nevion@gmail.com> // 2016, 2021 Costantino Grana <costantino.grana@unimore.it> // 2016, 2021 Federico Bolelli <federico.bolelli@unimore.it> // 2016 Lorenzo Baraldi <lorenzo.baraldi@unimore.it> // 2016 Roberto Vezzani <roberto.vezzani@unimore.it> // 2016 Michele Cancilla <cancilla.michele@gmail.com> // 2021 Stefano Allegretti <stefano.allegretti@unimore.it> //M*/ // #include "precomp.hpp" #include <vector> namespace cv{ namespace connectedcomponents{ struct NoOp{ NoOp(){ } inline void init(int /*labels*/){ } inline void initElement(const int /*nlabels*/){ } inline void operator()(int r, int c, int l){ CV_UNUSED(r); CV_UNUSED(c); CV_UNUSED(l); } void finish(){ } inline void setNextLoc(const int /*nextLoc*/){ } inline static void mergeStats(const cv::Mat& /*imgLabels*/, NoOp * /*sopArray*/, NoOp& /*sop*/, const int& /*nLabels*/){ } }; struct Point2ui64{ uint64 x, y; Point2ui64(uint64 _x, uint64 _y) :x(_x), y(_y){} }; struct CCStatsOp{ const _OutputArray *_mstatsv; cv::Mat statsv; const _OutputArray *_mcentroidsv; cv::Mat centroidsv; std::vector<Point2ui64> integrals; int _nextLoc; CCStatsOp() : _mstatsv(0), _mcentroidsv(0), _nextLoc(0) {} CCStatsOp(OutputArray _statsv, OutputArray _centroidsv) : _mstatsv(&_statsv), _mcentroidsv(&_centroidsv), _nextLoc(0){} inline void init(int nlabels){ _mstatsv->create(cv::Size(CC_STAT_MAX, nlabels), cv::DataType<int>::type); statsv = _mstatsv->getMat(); _mcentroidsv->create(cv::Size(2, nlabels), cv::DataType<double>::type); centroidsv = _mcentroidsv->getMat(); for (int l = 0; l < (int)nlabels; ++l){ int *row = (int *)&statsv.at<int>(l, 0); row[CC_STAT_LEFT] = INT_MAX; row[CC_STAT_TOP] = INT_MAX; row[CC_STAT_WIDTH] = INT_MIN; row[CC_STAT_HEIGHT] = INT_MIN; row[CC_STAT_AREA] = 0; } integrals.resize(nlabels, Point2ui64(0, 0)); } inline void initElement(const int nlabels){ statsv = cv::Mat(nlabels, CC_STAT_MAX, cv::DataType<int>::type); for (int l = 0; l < (int)nlabels; ++l){ int *row = (int *)statsv.ptr(l); row[CC_STAT_LEFT] = INT_MAX; row[CC_STAT_TOP] = INT_MAX; row[CC_STAT_WIDTH] = INT_MIN; row[CC_STAT_HEIGHT] = INT_MIN; row[CC_STAT_AREA] = 0; } integrals.resize(nlabels, Point2ui64(0, 0)); } void operator()(int r, int c, int l){ int *row =& statsv.at<int>(l, 0); row[CC_STAT_LEFT] = MIN(row[CC_STAT_LEFT], c); row[CC_STAT_WIDTH] = MAX(row[CC_STAT_WIDTH], c); row[CC_STAT_TOP] = MIN(row[CC_STAT_TOP], r); row[CC_STAT_HEIGHT] = MAX(row[CC_STAT_HEIGHT], r); row[CC_STAT_AREA]++; Point2ui64& integral = integrals[l]; integral.x += c; integral.y += r; } void finish(){ for (int l = 0; l < statsv.rows; ++l){ int *row =& statsv.at<int>(l, 0); double area = ((unsigned*)row)[CC_STAT_AREA]; double *centroid = ¢roidsv.at<double>(l, 0); if (area > 0){ row[CC_STAT_WIDTH] = row[CC_STAT_WIDTH] - row[CC_STAT_LEFT] + 1; row[CC_STAT_HEIGHT] = row[CC_STAT_HEIGHT] - row[CC_STAT_TOP] + 1; Point2ui64& integral = integrals[l]; centroid[0] = double(integral.x) / area; centroid[1] = double(integral.y) / area; } else { row[CC_STAT_WIDTH] = 0; row[CC_STAT_HEIGHT] = 0; row[CC_STAT_LEFT] = -1; centroid[0] = std::numeric_limits<double>::quiet_NaN(); centroid[1] = std::numeric_limits<double>::quiet_NaN(); } } } inline void setNextLoc(const int nextLoc){ _nextLoc = nextLoc; } inline static void mergeStats(const cv::Mat& imgLabels, CCStatsOp *sopArray, CCStatsOp& sop, const int& nLabels){ const int h = imgLabels.rows; if (sop._nextLoc != h){ for (int nextLoc = sop._nextLoc; nextLoc < h; nextLoc = sopArray[nextLoc]._nextLoc){ //merge between sopNext and sop for (int l = 0; l < nLabels; ++l){ int *rowNext = (int*)sopArray[nextLoc].statsv.ptr(l); if (rowNext[CC_STAT_AREA] > 0){ //if changed merge all the stats int *rowMerged = (int*)sop.statsv.ptr(l); rowMerged[CC_STAT_LEFT] = MIN(rowMerged[CC_STAT_LEFT], rowNext[CC_STAT_LEFT]); rowMerged[CC_STAT_WIDTH] = MAX(rowMerged[CC_STAT_WIDTH], rowNext[CC_STAT_WIDTH]); rowMerged[CC_STAT_TOP] = MIN(rowMerged[CC_STAT_TOP], rowNext[CC_STAT_TOP]); rowMerged[CC_STAT_HEIGHT] = MAX(rowMerged[CC_STAT_HEIGHT], rowNext[CC_STAT_HEIGHT]); rowMerged[CC_STAT_AREA] += rowNext[CC_STAT_AREA]; sop.integrals[l].x += sopArray[nextLoc].integrals[l].x; sop.integrals[l].y += sopArray[nextLoc].integrals[l].y; } } } } } }; //Find the root of the tree of node i template<typename LabelT> inline static LabelT findRoot(const LabelT *P, LabelT i){ LabelT root = i; while (P[root] < root){ root = P[root]; } return root; } //Make all nodes in the path of node i point to root template<typename LabelT> inline static void setRoot(LabelT *P, LabelT i, LabelT root){ while (P[i] < i){ LabelT j = P[i]; P[i] = root; i = j; } P[i] = root; } //Find the root of the tree of the node i and compress the path in the process template<typename LabelT> inline static LabelT find(LabelT *P, LabelT i){ LabelT root = findRoot(P, i); setRoot(P, i, root); return root; } //unite the two trees containing nodes i and j and return the new root template<typename LabelT> inline static LabelT set_union(LabelT *P, LabelT i, LabelT j){ LabelT root = findRoot(P, i); if (i != j){ LabelT rootj = findRoot(P, j); if (root > rootj){ root = rootj; } setRoot(P, j, root); } setRoot(P, i, root); return root; } //Flatten the Union Find tree and relabel the components template<typename LabelT> inline static LabelT flattenL(LabelT *P, LabelT length){ LabelT k = 1; for (LabelT i = 1; i < length; ++i){ if (P[i] < i){ P[i] = P[P[i]]; } else{ P[i] = k; k = k + 1; } } return k; } template<typename LabelT> inline static void flattenL(LabelT *P, const int start, const int nElem, LabelT& k){ for (int i = start; i < start + nElem; ++i){ if (P[i] < i){//node that point to root P[i] = P[P[i]]; } else{ //for root node P[i] = k; k = k + 1; } } } template <typename LT> static inline LT stripeFirstLabel4Connectivity(int y, int w) { CV_DbgAssert((y & 1) == 0); return (LT(y) * LT(w) /*+ 1*/) / 2 + 1; } template <typename LT> static inline LT stripeFirstLabel8Connectivity(int y, int w) { CV_DbgAssert((y & 1) == 0); return LT((y /*+ 1*/) / 2) * LT((w + 1) / 2) + 1; } //Parallel implementation of Spaghetti algorithm, described in "Spaghetti Labeling: Directed Acyclic Graphs //for Block-Based Connected Components Labeling", IEEE Transactions on Image Processing, Federico Bolelli et. al. //Parallelization method described in "Two More Strategies to Speed Up Connected Components Labeling Algorithms", //Image Analysis and Processing - ICIAP 2017, Federico Bolelli et. al. template<typename LabelT, typename PixelT, typename StatsOp = NoOp > struct LabelingBolelliParallel { class FirstScan : public cv::ParallelLoopBody { private: const cv::Mat& img_; cv::Mat& imgLabels_; LabelT* P_; int* chunksSizeAndLabels_; public: FirstScan(const cv::Mat& img, cv::Mat& imgLabels, LabelT* P, int* chunksSizeAndLabels) : img_(img), imgLabels_(imgLabels), P_(P), chunksSizeAndLabels_(chunksSizeAndLabels) {} FirstScan& operator=(const FirstScan&) { return *this; } void operator()(const cv::Range& range2) const CV_OVERRIDE { const Range range(range2.start * 2, std::min(range2.end * 2, img_.rows)); const int startR = range.start; chunksSizeAndLabels_[startR] = range.end; LabelT label = stripeFirstLabel8Connectivity<LabelT>(startR, imgLabels_.cols); const LabelT firstLabel = label; //const int h = img_.rows; const int w = img_.cols; const int stripe_h = range.end - range.start; //const int limitLine = startR + 1; const int e_rows = stripe_h & -2; const bool o_rows = stripe_h % 2 == 1; //const int e_cols = w & -2; //const bool o_cols = w % 2 == 1; { #define CONDITION_B img_row_prev_prev[c-1]>0 #define CONDITION_C img_row_prev_prev[c]>0 #define CONDITION_D img_row_prev_prev[c+1]>0 #define CONDITION_E img_row_prev_prev[c+2]>0 #define CONDITION_G img_row_prev[c-2]>0 #define CONDITION_H img_row_prev[c-1]>0 #define CONDITION_I img_row_prev[c]>0 #define CONDITION_J img_row_prev[c+1]>0 #define CONDITION_K img_row_prev[c+2]>0 #define CONDITION_M img_row[c-2]>0 #define CONDITION_N img_row[c-1]>0 #define CONDITION_O img_row[c]>0 #define CONDITION_P img_row[c+1]>0 #define CONDITION_R img_row_fol[c-1]>0 #define CONDITION_S img_row_fol[c]>0 #define CONDITION_T img_row_fol[c+1]>0 // Action 1: No action #define ACTION_1 img_labels_row[c] = 0; // Action 2: New label (the block has foreground pixels and is not connected to anything else) #define ACTION_2 img_labels_row[c] = label; \ P_[label] = label; \ label = label + 1; //Action 3: Assign label of block P #define ACTION_3 img_labels_row[c] = img_labels_row_prev_prev[c - 2]; // Action 4: Assign label of block Q #define ACTION_4 img_labels_row[c] = img_labels_row_prev_prev[c]; // Action 5: Assign label of block R #define ACTION_5 img_labels_row[c] = img_labels_row_prev_prev[c + 2]; // Action 6: Assign label of block S #define ACTION_6 img_labels_row[c] = img_labels_row[c - 2]; // Action 7: Merge labels of block P and Q #define ACTION_7 img_labels_row[c] = set_union(P_, img_labels_row_prev_prev[c - 2], img_labels_row_prev_prev[c]); //Action 8: Merge labels of block P and R #define ACTION_8 img_labels_row[c] = set_union(P_, img_labels_row_prev_prev[c - 2], img_labels_row_prev_prev[c + 2]); // Action 9 Merge labels of block P and S #define ACTION_9 img_labels_row[c] = set_union(P_, img_labels_row_prev_prev[c - 2], img_labels_row[c - 2]); // Action 10 Merge labels of block Q and R #define ACTION_10 img_labels_row[c] = set_union(P_, img_labels_row_prev_prev[c], img_labels_row_prev_prev[c + 2]); // Action 11: Merge labels of block Q and S #define ACTION_11 img_labels_row[c] = set_union(P_, img_labels_row_prev_prev[c], img_labels_row[c - 2]); // Action 12: Merge labels of block R and S #define ACTION_12 img_labels_row[c] = set_union(P_, img_labels_row_prev_prev[c + 2], img_labels_row[c - 2]); // Action 13: Merge labels of block P, Q and R #define ACTION_13 img_labels_row[c] = set_union(P_, set_union(P_, img_labels_row_prev_prev[c - 2], img_labels_row_prev_prev[c]), img_labels_row_prev_prev[c + 2]); // Action 14: Merge labels of block P, Q and S #define ACTION_14 img_labels_row[c] = set_union(P_, set_union(P_, img_labels_row_prev_prev[c - 2], img_labels_row_prev_prev[c]), img_labels_row[c - 2]); //Action 15: Merge labels of block P, R and S #define ACTION_15 img_labels_row[c] = set_union(P_, set_union(P_, img_labels_row_prev_prev[c - 2], img_labels_row_prev_prev[c + 2]), img_labels_row[c - 2]); //Action 16: labels of block Q, R and S #define ACTION_16 img_labels_row[c] = set_union(P_, set_union(P_, img_labels_row_prev_prev[c], img_labels_row_prev_prev[c + 2]), img_labels_row[c - 2]); } // The following Directed Rooted Acyclic Graphs (DAGs) allow to choose which action to // perform, checking as few conditions as possible. Special DAGs are used for the first/last // line of the image and for single line images. Actions: the blocks label are provisionally // stored in the top left pixel of the block in the labels image. if (stripe_h == 1) { // Single line const PixelT* const img_row = img_.ptr<PixelT>(startR); LabelT* const img_labels_row = imgLabels_.ptr<LabelT>(startR); int c = -2; #include "ccl_bolelli_forest_singleline.inc.hpp" } else { // More than one line // First couple of lines { const PixelT* const img_row = img_.ptr<PixelT>(startR); const PixelT* const img_row_fol = (PixelT*)(((char*)img_row) + img_.step.p[0]); LabelT* const img_labels_row = imgLabels_.ptr<LabelT>(startR); int c = -2; #include "ccl_bolelli_forest_firstline.inc.hpp" } // Every other line but the last one if image has an odd number of rows for (int r = startR + 2; r < startR + e_rows; r += 2) { // Get rows pointer const PixelT* const img_row = img_.ptr<PixelT>(r); const PixelT* const img_row_prev = (PixelT*)(((char*)img_row) - img_.step.p[0]); const PixelT* const img_row_prev_prev = (PixelT*)(((char*)img_row_prev) - img_.step.p[0]); const PixelT* const img_row_fol = (PixelT*)(((char*)img_row) + img_.step.p[0]); LabelT* const img_labels_row = imgLabels_.ptr<LabelT>(r); LabelT* const img_labels_row_prev_prev = (LabelT*)(((char*)img_labels_row) - imgLabels_.step.p[0] - imgLabels_.step.p[0]); int c = -2; goto tree_0; #include "ccl_bolelli_forest.inc.hpp" } // Last line (in case the rows are odd) if (o_rows) { const int r = startR + stripe_h - 1; const PixelT* const img_row = img_.ptr<PixelT>(r); const PixelT* const img_row_prev = (PixelT*)(((char*)img_row) - img_.step.p[0]); const PixelT* const img_row_prev_prev = (PixelT*)(((char*)img_row_prev) - img_.step.p[0]); LabelT* const img_labels_row = imgLabels_.ptr<LabelT>(r); LabelT* const img_labels_row_prev_prev = (LabelT*)(((char*)img_labels_row) - imgLabels_.step.p[0] - imgLabels_.step.p[0]); int c = -2; #include "ccl_bolelli_forest_lastline.inc.hpp" } } //write in the follower memory location chunksSizeAndLabels_[startR + 1] = label - firstLabel; // undef conditions and actions { #undef ACTION_1 #undef ACTION_2 #undef ACTION_3 #undef ACTION_4 #undef ACTION_5 #undef ACTION_6 #undef ACTION_7 #undef ACTION_8 #undef ACTION_9 #undef ACTION_10 #undef ACTION_11 #undef ACTION_12 #undef ACTION_13 #undef ACTION_14 #undef ACTION_15 #undef ACTION_16 #undef CONDITION_B #undef CONDITION_C #undef CONDITION_D #undef CONDITION_E #undef CONDITION_G #undef CONDITION_H #undef CONDITION_I #undef CONDITION_J #undef CONDITION_K #undef CONDITION_M #undef CONDITION_N #undef CONDITION_O #undef CONDITION_P #undef CONDITION_R #undef CONDITION_S #undef CONDITION_T } } }; class SecondScan : public cv::ParallelLoopBody { private: const cv::Mat& img_; cv::Mat& imgLabels_; LabelT* P_; StatsOp& sop_; StatsOp* sopArray_; LabelT& nLabels_; public: SecondScan(const cv::Mat& img, cv::Mat& imgLabels, LabelT* P, StatsOp& sop, StatsOp* sopArray, LabelT& nLabels) : img_(img), imgLabels_(imgLabels), P_(P), sop_(sop), sopArray_(sopArray), nLabels_(nLabels) {} void operator()(const cv::Range& range2) const CV_OVERRIDE { const Range range(range2.start * 2, std::min(range2.end * 2, img_.rows)); int r = range.start; const int rowBegin = r; const int rowEnd = range.end; if (rowBegin > 0) { sopArray_[rowBegin].initElement(nLabels_); sopArray_[rowBegin].setNextLoc(rowEnd); //_nextLoc = rowEnd; if (imgLabels_.rows & 1) { if (imgLabels_.cols & 1) { //Case 1: both rows and cols odd for (; r < rowEnd; r += 2) { // Get rows pointer const PixelT* const img_row = img_.ptr<PixelT>(r); const PixelT* const img_row_fol = (PixelT*)(((char*)img_row) + img_.step.p[0]); LabelT* const imgLabels_row = imgLabels_.ptr<LabelT>(r); LabelT* const imgLabels_row_fol = (LabelT*)(((char*)imgLabels_row) + imgLabels_.step.p[0]); // Get rows pointer for (int c = 0; c < imgLabels_.cols; c += 2) { LabelT iLabel = imgLabels_row[c]; if (iLabel > 0) { iLabel = P_[iLabel]; if (img_row[c] > 0) { imgLabels_row[c] = iLabel; sopArray_[rowBegin](r, c, iLabel); } else { imgLabels_row[c] = 0; sopArray_[rowBegin](r, c, 0); } if (c + 1 < imgLabels_.cols) { if (img_row[c + 1] > 0) { imgLabels_row[c + 1] = iLabel; sopArray_[rowBegin](r, c + 1, iLabel); } else { imgLabels_row[c + 1] = 0; sopArray_[rowBegin](r, c + 1, 0); } if (r + 1 < imgLabels_.rows) { if (img_row_fol[c] > 0) { imgLabels_row_fol[c] = iLabel; sopArray_[rowBegin](r + 1, c, iLabel); } else { imgLabels_row_fol[c] = 0; sopArray_[rowBegin](r + 1, c, 0); } if (img_row_fol[c + 1] > 0) { imgLabels_row_fol[c + 1] = iLabel; sopArray_[rowBegin](r + 1, c + 1, iLabel); } else { imgLabels_row_fol[c + 1] = 0; sopArray_[rowBegin](r + 1, c + 1, 0); } } } else if (r + 1 < imgLabels_.rows) { if (img_row_fol[c] > 0) { imgLabels_row_fol[c] = iLabel; sopArray_[rowBegin](r + 1, c, iLabel); } else { imgLabels_row_fol[c] = 0; sopArray_[rowBegin](r + 1, c, 0); } } } else { imgLabels_row[c] = 0; sopArray_[rowBegin](r, c, 0); if (c + 1 < imgLabels_.cols) { imgLabels_row[c + 1] = 0; sopArray_[rowBegin](r, c + 1, 0); if (r + 1 < imgLabels_.rows) { imgLabels_row_fol[c] = 0; imgLabels_row_fol[c + 1] = 0; sopArray_[rowBegin](r + 1, c, 0); sopArray_[rowBegin](r + 1, c + 1, 0); } } else if (r + 1 < imgLabels_.rows) { imgLabels_row_fol[c] = 0; sopArray_[rowBegin](r + 1, c, 0); } } } } }//END Case 1 else { //Case 2: only rows odd for (; r < rowEnd; r += 2) { // Get rows pointer const PixelT* const img_row = img_.ptr<PixelT>(r); const PixelT* const img_row_fol = (PixelT*)(((char*)img_row) + img_.step.p[0]); LabelT* const imgLabels_row = imgLabels_.ptr<LabelT>(r); LabelT* const imgLabels_row_fol = (LabelT*)(((char*)imgLabels_row) + imgLabels_.step.p[0]); // Get rows pointer for (int c = 0; c < imgLabels_.cols; c += 2) { LabelT iLabel = imgLabels_row[c]; if (iLabel > 0) { iLabel = P_[iLabel]; if (img_row[c] > 0) { imgLabels_row[c] = iLabel; sopArray_[rowBegin](r, c, iLabel); } else { imgLabels_row[c] = 0; sopArray_[rowBegin](r, c, 0); } if (img_row[c + 1] > 0) { imgLabels_row[c + 1] = iLabel; sopArray_[rowBegin](r, c + 1, iLabel); } else { imgLabels_row[c + 1] = 0; sopArray_[rowBegin](r, c + 1, 0); } if (r + 1 < imgLabels_.rows) { if (img_row_fol[c] > 0) { imgLabels_row_fol[c] = iLabel; sopArray_[rowBegin](r + 1, c, iLabel); } else { imgLabels_row_fol[c] = 0; sopArray_[rowBegin](r + 1, c, 0); } if (img_row_fol[c + 1] > 0) { imgLabels_row_fol[c + 1] = iLabel; sopArray_[rowBegin](r + 1, c + 1, iLabel); } else { imgLabels_row_fol[c + 1] = 0; sopArray_[rowBegin](r + 1, c + 1, 0); } } } else { imgLabels_row[c] = 0; imgLabels_row[c + 1] = 0; sopArray_[rowBegin](r, c, 0); sopArray_[rowBegin](r, c + 1, 0); if (r + 1 < imgLabels_.rows) { imgLabels_row_fol[c] = 0; imgLabels_row_fol[c + 1] = 0; sopArray_[rowBegin](r + 1, c, 0); sopArray_[rowBegin](r + 1, c + 1, 0); } } } } }// END Case 2 } else { if (imgLabels_.cols & 1) { //Case 3: only cols odd for (; r < rowEnd; r += 2) { // Get rows pointer const PixelT* const img_row = img_.ptr<PixelT>(r); const PixelT* const img_row_fol = (PixelT*)(((char*)img_row) + img_.step.p[0]); LabelT* const imgLabels_row = imgLabels_.ptr<LabelT>(r); LabelT* const imgLabels_row_fol = (LabelT*)(((char*)imgLabels_row) + imgLabels_.step.p[0]); // Get rows pointer for (int c = 0; c < imgLabels_.cols; c += 2) { LabelT iLabel = imgLabels_row[c]; if (iLabel > 0) { iLabel = P_[iLabel]; if (img_row[c] > 0) { imgLabels_row[c] = iLabel; sopArray_[rowBegin](r, c, iLabel); } else { imgLabels_row[c] = 0; sopArray_[rowBegin](r, c, 0); } if (img_row_fol[c] > 0) { imgLabels_row_fol[c] = iLabel; sopArray_[rowBegin](r + 1, c, iLabel); } else { imgLabels_row_fol[c] = 0; sopArray_[rowBegin](r + 1, c, 0); } if (c + 1 < imgLabels_.cols) { if (img_row[c + 1] > 0) { imgLabels_row[c + 1] = iLabel; sopArray_[rowBegin](r, c + 1, iLabel); } else { imgLabels_row[c + 1] = 0; sopArray_[rowBegin](r, c + 1, 0); } if (img_row_fol[c + 1] > 0) { imgLabels_row_fol[c + 1] = iLabel; sopArray_[rowBegin](r + 1, c + 1, iLabel); } else { imgLabels_row_fol[c + 1] = 0; sopArray_[rowBegin](r + 1, c + 1, 0); } } } else { imgLabels_row[c] = 0; imgLabels_row_fol[c] = 0; sopArray_[rowBegin](r, c, 0); sopArray_[rowBegin](r + 1, c, 0); if (c + 1 < imgLabels_.cols) { imgLabels_row[c + 1] = 0; imgLabels_row_fol[c + 1] = 0; sopArray_[rowBegin](r, c + 1, 0); sopArray_[rowBegin](r + 1, c + 1, 0); } } } } }// END case 3 else { //Case 4: nothing odd for (; r < rowEnd; r += 2) { // Get rows pointer const PixelT* const img_row = img_.ptr<PixelT>(r); const PixelT* const img_row_fol = (PixelT*)(((char*)img_row) + img_.step.p[0]); LabelT* const imgLabels_row = imgLabels_.ptr<LabelT>(r); LabelT* const imgLabels_row_fol = (LabelT*)(((char*)imgLabels_row) + imgLabels_.step.p[0]); // Get rows pointer for (int c = 0; c < imgLabels_.cols; c += 2) { LabelT iLabel = imgLabels_row[c]; if (iLabel > 0) { iLabel = P_[iLabel]; if (img_row[c] > 0) { imgLabels_row[c] = iLabel; sopArray_[rowBegin](r, c, iLabel); } else { imgLabels_row[c] = 0; sopArray_[rowBegin](r, c, 0); } if (img_row[c + 1] > 0) { imgLabels_row[c + 1] = iLabel; sopArray_[rowBegin](r, c + 1, iLabel); } else { imgLabels_row[c + 1] = 0; sopArray_[rowBegin](r, c + 1, 0); } if (img_row_fol[c] > 0) { imgLabels_row_fol[c] = iLabel; sopArray_[rowBegin](r + 1, c, iLabel); } else { imgLabels_row_fol[c] = 0; sopArray_[rowBegin](r + 1, c, 0); } if (img_row_fol[c + 1] > 0) { imgLabels_row_fol[c + 1] = iLabel; sopArray_[rowBegin](r + 1, c + 1, iLabel); } else { imgLabels_row_fol[c + 1] = 0; sopArray_[rowBegin](r + 1, c + 1, 0); } } else { imgLabels_row[c] = 0; imgLabels_row[c + 1] = 0; imgLabels_row_fol[c] = 0; imgLabels_row_fol[c + 1] = 0; sopArray_[rowBegin](r, c, 0); sopArray_[rowBegin](r, c + 1, 0); sopArray_[rowBegin](r + 1, c, 0); sopArray_[rowBegin](r + 1, c + 1, 0); } } }//END case 4 } } } else { //the first thread uses sop in order to make less merges sop_.setNextLoc(rowEnd); if (imgLabels_.rows & 1) { if (imgLabels_.cols & 1) { //Case 1: both rows and cols odd for (; r < rowEnd; r += 2) { // Get rows pointer const PixelT* const img_row = img_.ptr<PixelT>(r); const PixelT* const img_row_fol = (PixelT*)(((char*)img_row) + img_.step.p[0]); LabelT* const imgLabels_row = imgLabels_.ptr<LabelT>(r); LabelT* const imgLabels_row_fol = (LabelT*)(((char*)imgLabels_row) + imgLabels_.step.p[0]); // Get rows pointer for (int c = 0; c < imgLabels_.cols; c += 2) { LabelT iLabel = imgLabels_row[c]; if (iLabel > 0) { iLabel = P_[iLabel]; if (img_row[c] > 0) { imgLabels_row[c] = iLabel; sop_(r, c, iLabel); } else { imgLabels_row[c] = 0; sop_(r, c, 0); } if (c + 1 < imgLabels_.cols) { if (img_row[c + 1] > 0) { imgLabels_row[c + 1] = iLabel; sop_(r, c + 1, iLabel); } else { imgLabels_row[c + 1] = 0; sop_(r, c + 1, 0); } if (r + 1 < imgLabels_.rows) { if (img_row_fol[c] > 0) { imgLabels_row_fol[c] = iLabel; sop_(r + 1, c, iLabel); } else { imgLabels_row_fol[c] = 0; sop_(r + 1, c, 0); } if (img_row_fol[c + 1] > 0) { imgLabels_row_fol[c + 1] = iLabel; sop_(r + 1, c + 1, iLabel); } else { imgLabels_row_fol[c + 1] = 0; sop_(r + 1, c + 1, 0); } } } else if (r + 1 < imgLabels_.rows) { if (img_row_fol[c] > 0) { imgLabels_row_fol[c] = iLabel; sop_(r + 1, c, iLabel); } else { imgLabels_row_fol[c] = 0; sop_(r + 1, c, 0); } } } else { imgLabels_row[c] = 0; sop_(r, c, 0); if (c + 1 < imgLabels_.cols) { imgLabels_row[c + 1] = 0; sop_(r, c + 1, 0); if (r + 1 < imgLabels_.rows) { imgLabels_row_fol[c] = 0; imgLabels_row_fol[c + 1] = 0; sop_(r + 1, c, 0); sop_(r + 1, c + 1, 0); } } else if (r + 1 < imgLabels_.rows) { imgLabels_row_fol[c] = 0; sop_(r + 1, c, 0); } } } } }//END Case 1 else { //Case 2: only rows odd for (; r < rowEnd; r += 2) { // Get rows pointer const PixelT* const img_row = img_.ptr<PixelT>(r); const PixelT* const img_row_fol = (PixelT*)(((char*)img_row) + img_.step.p[0]); LabelT* const imgLabels_row = imgLabels_.ptr<LabelT>(r); LabelT* const imgLabels_row_fol = (LabelT*)(((char*)imgLabels_row) + imgLabels_.step.p[0]); // Get rows pointer for (int c = 0; c < imgLabels_.cols; c += 2) { LabelT iLabel = imgLabels_row[c]; if (iLabel > 0) { iLabel = P_[iLabel]; if (img_row[c] > 0) { imgLabels_row[c] = iLabel; sop_(r, c, iLabel); } else { imgLabels_row[c] = 0; sop_(r, c, 0); } if (img_row[c + 1] > 0) { imgLabels_row[c + 1] = iLabel; sop_(r, c + 1, iLabel); } else { imgLabels_row[c + 1] = 0; sop_(r, c + 1, 0); } if (r + 1 < imgLabels_.rows) { if (img_row_fol[c] > 0) { imgLabels_row_fol[c] = iLabel; sop_(r + 1, c, iLabel); } else { imgLabels_row_fol[c] = 0; sop_(r + 1, c, 0); } if (img_row_fol[c + 1] > 0) { imgLabels_row_fol[c + 1] = iLabel; sop_(r + 1, c + 1, iLabel); } else { imgLabels_row_fol[c + 1] = 0; sop_(r + 1, c + 1, 0); } } } else { imgLabels_row[c] = 0; imgLabels_row[c + 1] = 0; sop_(r, c, 0); sop_(r, c + 1, 0); if (r + 1 < imgLabels_.rows) { imgLabels_row_fol[c] = 0; imgLabels_row_fol[c + 1] = 0; sop_(r + 1, c, 0); sop_(r + 1, c + 1, 0); } } } } }// END Case 2 } else { if (imgLabels_.cols & 1) { //Case 3: only cols odd for (; r < rowEnd; r += 2) { // Get rows pointer const PixelT* const img_row = img_.ptr<PixelT>(r); const PixelT* const img_row_fol = (PixelT*)(((char*)img_row) + img_.step.p[0]); LabelT* const imgLabels_row = imgLabels_.ptr<LabelT>(r); LabelT* const imgLabels_row_fol = (LabelT*)(((char*)imgLabels_row) + imgLabels_.step.p[0]); // Get rows pointer for (int c = 0; c < imgLabels_.cols; c += 2) { LabelT iLabel = imgLabels_row[c]; if (iLabel > 0) { iLabel = P_[iLabel]; if (img_row[c] > 0) { imgLabels_row[c] = iLabel; sop_(r, c, iLabel); } else { imgLabels_row[c] = 0; sop_(r, c, 0); } if (img_row_fol[c] > 0) { imgLabels_row_fol[c] = iLabel; sop_(r + 1, c, iLabel); } else { imgLabels_row_fol[c] = 0; sop_(r + 1, c, 0); } if (c + 1 < imgLabels_.cols) { if (img_row[c + 1] > 0) { imgLabels_row[c + 1] = iLabel; sop_(r, c + 1, iLabel); } else { imgLabels_row[c + 1] = 0; sop_(r, c + 1, 0); } if (img_row_fol[c + 1] > 0) { imgLabels_row_fol[c + 1] = iLabel; sop_(r + 1, c + 1, iLabel); } else { imgLabels_row_fol[c + 1] = 0; sop_(r + 1, c + 1, 0); } } } else { imgLabels_row[c] = 0; imgLabels_row_fol[c] = 0; sop_(r, c, 0); sop_(r + 1, c, 0); if (c + 1 < imgLabels_.cols) { imgLabels_row[c + 1] = 0; imgLabels_row_fol[c + 1] = 0; sop_(r, c + 1, 0); sop_(r + 1, c + 1, 0); } } } } }// END case 3 else { //Case 4: nothing odd for (; r < rowEnd; r += 2) { // Get rows pointer const PixelT* const img_row = img_.ptr<PixelT>(r); const PixelT* const img_row_fol = (PixelT*)(((char*)img_row) + img_.step.p[0]); LabelT* const imgLabels_row = imgLabels_.ptr<LabelT>(r); LabelT* const imgLabels_row_fol = (LabelT*)(((char*)imgLabels_row) + imgLabels_.step.p[0]); // Get rows pointer for (int c = 0; c < imgLabels_.cols; c += 2) { LabelT iLabel = imgLabels_row[c]; if (iLabel > 0) { iLabel = P_[iLabel]; if (img_row[c] > 0) { imgLabels_row[c] = iLabel; sop_(r, c, iLabel); } else { imgLabels_row[c] = 0; sop_(r, c, 0); } if (img_row[c + 1] > 0) { imgLabels_row[c + 1] = iLabel; sop_(r, c + 1, iLabel); } else { imgLabels_row[c + 1] = 0; sop_(r, c + 1, 0); } if (img_row_fol[c] > 0) { imgLabels_row_fol[c] = iLabel; sop_(r + 1, c, iLabel); } else { imgLabels_row_fol[c] = 0; sop_(r + 1, c, 0); } if (img_row_fol[c + 1] > 0) { imgLabels_row_fol[c + 1] = iLabel; sop_(r + 1, c + 1, iLabel); } else { imgLabels_row_fol[c + 1] = 0; sop_(r + 1, c + 1, 0); } } else { imgLabels_row[c] = 0; imgLabels_row[c + 1] = 0; imgLabels_row_fol[c] = 0; imgLabels_row_fol[c + 1] = 0; sop_(r, c, 0); sop_(r, c + 1, 0); sop_(r + 1, c, 0); sop_(r + 1, c + 1, 0); } } }//END case 4 } } } } }; inline static void mergeLabels(const cv::Mat& img, cv::Mat& imgLabels, LabelT* P, int* chunksSizeAndLabels) { // Merge Mask // +---+---+---+ // |P -|Q -|R -| // |- -|- -|- -| // +---+---+---+ // |X -| // |- -| // +---+ const int w = imgLabels.cols, h = imgLabels.rows; for (int r = chunksSizeAndLabels[0]; r < h; r = chunksSizeAndLabels[r]) { LabelT* const imgLabels_row = imgLabels.ptr<LabelT>(r); LabelT* const imgLabels_row_prev_prev = (LabelT*)(((char*)imgLabels_row) - imgLabels.step.p[0] - imgLabels.step.p[0]); const PixelT* const img_row = img.ptr<PixelT>(r); const PixelT* const img_row_prev = (PixelT*)(((char*)img_row) - img.step.p[0]); for (int c = 0; c < w; c += 2) { #define condition_x imgLabels_row[c] > 0 #define condition_pppr c > 1 && imgLabels_row_prev_prev[c - 2] > 0 #define condition_qppr imgLabels_row_prev_prev[c] > 0 #define condition_qppr1 c < w - 1 #define condition_qppr2 c < w #define condition_rppr c < w - 2 && imgLabels_row_prev_prev[c + 2] > 0 if (condition_x) { if (condition_pppr) { //check in img if (img_row[c] > 0 && img_row_prev[c - 1] > 0) //assign the same label imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c - 2], imgLabels_row[c]); } if (condition_qppr) { if (condition_qppr1) { if ((img_row[c] > 0 && img_row_prev[c] > 0) || (img_row[c + 1] > 0 && img_row_prev[c] > 0) || (img_row[c] > 0 && img_row_prev[c + 1] > 0) || (img_row[c + 1] > 0 && img_row_prev[c + 1] > 0)) { imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c]); } } else /*if (condition_qppr2)*/ { if (img_row[c] > 0 && img_row_prev[c] > 0) imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c]); } } if (condition_rppr) { if (img_row[c + 1] > 0 && img_row_prev[c + 2] > 0) imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c]); } } #undef condition_x #undef condition_pppr #undef condition_qppr #undef condition_qppr1 #undef condition_qppr2 #undef condition_rppr } } } LabelT operator()(const cv::Mat& img, cv::Mat& imgLabels, int connectivity, StatsOp& sop) { CV_Assert(img.rows == imgLabels.rows); CV_Assert(img.cols == imgLabels.cols); CV_Assert(connectivity == 8); const int h = img.rows; const int w = img.cols; //A quick and dirty upper bound for the maximum number of labels. //Following formula comes from the fact that a 2x2 block in 8-connectivity case //can never have more than 1 new label and 1 label for background. //Worst case image example pattern: //1 0 1 0 1... //0 0 0 0 0... //1 0 1 0 1... //............ const size_t Plength = size_t(((h + 1) / 2) * size_t((w + 1) / 2)) + 1; //Array used to store info and labeled pixel by each thread. //Different threads affect different memory location of chunksSizeAndLabels const int chunksSizeAndLabelsSize = roundUp(h, 2); std::vector<int> chunksSizeAndLabels(chunksSizeAndLabelsSize); //Tree of labels std::vector<LabelT> P(Plength, 0); //First label is for background //P[0] = 0; cv::Range range2(0, divUp(h, 2)); const double nParallelStripes = std::max(1, std::min(h / 2, getNumThreads() * 4)); //First scan cv::parallel_for_(range2, FirstScan(img, imgLabels, P.data(), chunksSizeAndLabels.data()), nParallelStripes); //merge labels of different chunks mergeLabels(img, imgLabels, P.data(), chunksSizeAndLabels.data()); LabelT nLabels = 1; for (int i = 0; i < h; i = chunksSizeAndLabels[i]) { CV_DbgAssert(i + 1 < chunksSizeAndLabelsSize); flattenL(P.data(), stripeFirstLabel8Connectivity<LabelT>(i, w), chunksSizeAndLabels[i + 1], nLabels); } //Array for statistics data std::vector<StatsOp> sopArray(h); sop.init(nLabels); //Second scan cv::parallel_for_(range2, SecondScan(img, imgLabels, P.data(), sop, sopArray.data(), nLabels), nParallelStripes); StatsOp::mergeStats(imgLabels, sopArray.data(), sop, nLabels); sop.finish(); return nLabels; } };//End struct LabelingBolelliParallel //Implementation of Spaghetti algorithm, as described in "Spaghetti Labeling: Directed Acyclic Graphs //for Block-Based Connected Components Labeling", IEEE Transactions on Image Processing, Federico Bolelli et. al. template<typename LabelT, typename PixelT, typename StatsOp = NoOp > struct LabelingBolelli { LabelT operator()(const cv::Mat& img, cv::Mat& imgLabels, int connectivity, StatsOp& sop) { CV_Assert(img.rows == imgLabels.rows); CV_Assert(img.cols == imgLabels.cols); CV_Assert(connectivity == 8); const int h = img.rows; const int w = img.cols; const int e_rows = h & -2; const bool o_rows = h % 2 == 1; const int e_cols = w & -2; const bool o_cols = w % 2 == 1; // A quick and dirty upper bound for the maximum number of labels. // Following formula comes from the fact that a 2x2 block in 8-connectivity case // can never have more than 1 new label and 1 label for background. // Worst case image example pattern: // 1 0 1 0 1... // 0 0 0 0 0... // 1 0 1 0 1... // ............ const size_t Plength = size_t(((h + 1) / 2) * size_t((w + 1) / 2)) + 1; std::vector<LabelT> P_(Plength, 0); LabelT *P = P_.data(); //P[0] = 0; LabelT lunique = 1; // First scan // We work with 2x2 blocks // +-+-+-+ // |P|Q|R| // +-+-+-+ // |S|X| // +-+-+ // The pixels are named as follows // +---+---+---+ // |a b|c d|e f| // |g h|i j|k l| // +---+---+---+ // |m n|o p| // |q r|s t| // +---+---+ // Pixels a, f, l, q are not needed, since we need to understand the // the connectivity between these blocks and those pixels only matter // when considering the outer connectivities // A bunch of defines is used to check if the pixels are foreground // and to define actions to be performed on blocks { #define CONDITION_B img_row_prev_prev[c-1]>0 #define CONDITION_C img_row_prev_prev[c]>0 #define CONDITION_D img_row_prev_prev[c+1]>0 #define CONDITION_E img_row_prev_prev[c+2]>0 #define CONDITION_G img_row_prev[c-2]>0 #define CONDITION_H img_row_prev[c-1]>0 #define CONDITION_I img_row_prev[c]>0 #define CONDITION_J img_row_prev[c+1]>0 #define CONDITION_K img_row_prev[c+2]>0 #define CONDITION_M img_row[c-2]>0 #define CONDITION_N img_row[c-1]>0 #define CONDITION_O img_row[c]>0 #define CONDITION_P img_row[c+1]>0 #define CONDITION_R img_row_fol[c-1]>0 #define CONDITION_S img_row_fol[c]>0 #define CONDITION_T img_row_fol[c+1]>0 // Action 1: No action #define ACTION_1 img_labels_row[c] = 0; // Action 2: New label (the block has foreground pixels and is not connected to anything else) #define ACTION_2 img_labels_row[c] = lunique; \ P[lunique] = lunique; \ lunique = lunique + 1; //Action 3: Assign label of block P #define ACTION_3 img_labels_row[c] = img_labels_row_prev_prev[c - 2]; // Action 4: Assign label of block Q #define ACTION_4 img_labels_row[c] = img_labels_row_prev_prev[c]; // Action 5: Assign label of block R #define ACTION_5 img_labels_row[c] = img_labels_row_prev_prev[c + 2]; // Action 6: Assign label of block S #define ACTION_6 img_labels_row[c] = img_labels_row[c - 2]; // Action 7: Merge labels of block P and Q #define ACTION_7 img_labels_row[c] = set_union(P, img_labels_row_prev_prev[c - 2], img_labels_row_prev_prev[c]); //Action 8: Merge labels of block P and R #define ACTION_8 img_labels_row[c] = set_union(P, img_labels_row_prev_prev[c - 2], img_labels_row_prev_prev[c + 2]); // Action 9 Merge labels of block P and S #define ACTION_9 img_labels_row[c] = set_union(P, img_labels_row_prev_prev[c - 2], img_labels_row[c - 2]); // Action 10 Merge labels of block Q and R #define ACTION_10 img_labels_row[c] = set_union(P, img_labels_row_prev_prev[c], img_labels_row_prev_prev[c + 2]); // Action 11: Merge labels of block Q and S #define ACTION_11 img_labels_row[c] = set_union(P, img_labels_row_prev_prev[c], img_labels_row[c - 2]); // Action 12: Merge labels of block R and S #define ACTION_12 img_labels_row[c] = set_union(P, img_labels_row_prev_prev[c + 2], img_labels_row[c - 2]); // Action 13: Merge labels of block P, Q and R #define ACTION_13 img_labels_row[c] = set_union(P, set_union(P, img_labels_row_prev_prev[c - 2], img_labels_row_prev_prev[c]), img_labels_row_prev_prev[c + 2]); // Action 14: Merge labels of block P, Q and S #define ACTION_14 img_labels_row[c] = set_union(P, set_union(P, img_labels_row_prev_prev[c - 2], img_labels_row_prev_prev[c]), img_labels_row[c - 2]); //Action 15: Merge labels of block P, R and S #define ACTION_15 img_labels_row[c] = set_union(P, set_union(P, img_labels_row_prev_prev[c - 2], img_labels_row_prev_prev[c + 2]), img_labels_row[c - 2]); //Action 16: labels of block Q, R and S #define ACTION_16 img_labels_row[c] = set_union(P, set_union(P, img_labels_row_prev_prev[c], img_labels_row_prev_prev[c + 2]), img_labels_row[c - 2]); } // The following Directed Rooted Acyclic Graphs (DAGs) allow to choose which action to // perform, checking as few conditions as possible. Special DAGs are used for the first/last // line of the image and for single line images. Actions: the blocks label are provisionally // stored in the top left pixel of the block in the labels image. if (h == 1) { // Single line const PixelT * const img_row = img.ptr<PixelT>(0); LabelT * const img_labels_row = imgLabels.ptr<LabelT>(0); int c = -2; #include "ccl_bolelli_forest_singleline.inc.hpp" } else { // More than one line // First couple of lines { const PixelT * const img_row = img.ptr<PixelT>(0); const PixelT * const img_row_fol = (PixelT *)(((char*)img_row) + img.step.p[0]); LabelT * const img_labels_row = imgLabels.ptr<LabelT>(0); int c = -2; #include "ccl_bolelli_forest_firstline.inc.hpp" } // Every other line but the last one if image has an odd number of rows for (int r = 2; r < e_rows; r += 2) { // Get rows pointer const PixelT * const img_row = img.ptr<PixelT>(r); const PixelT * const img_row_prev = (PixelT *)(((char*)img_row) - img.step.p[0]); const PixelT * const img_row_prev_prev = (PixelT *)(((char*)img_row_prev) - img.step.p[0]); const PixelT * const img_row_fol = (PixelT *)(((char*)img_row) + img.step.p[0]); LabelT * const img_labels_row = imgLabels.ptr<LabelT>(r); LabelT * const img_labels_row_prev_prev = (LabelT *)(((char*)img_labels_row) - imgLabels.step.p[0] - imgLabels.step.p[0]); int c = -2; goto tree_0; #include "ccl_bolelli_forest.inc.hpp" } // Last line (in case the rows are odd) if (o_rows) { int r = h - 1; const PixelT * const img_row = img.ptr<PixelT>(r); const PixelT * const img_row_prev = (PixelT *)(((char*)img_row) - img.step.p[0]); const PixelT * const img_row_prev_prev = (PixelT *)(((char*)img_row_prev) - img.step.p[0]); LabelT * const img_labels_row = imgLabels.ptr<LabelT>(r); LabelT * const img_labels_row_prev_prev = (LabelT *)(((char*)img_labels_row) - imgLabels.step.p[0] - imgLabels.step.p[0]); int c = -2; #include "ccl_bolelli_forest_lastline.inc.hpp" } } // undef conditions and actions { #undef ACTION_1 #undef ACTION_2 #undef ACTION_3 #undef ACTION_4 #undef ACTION_5 #undef ACTION_6 #undef ACTION_7 #undef ACTION_8 #undef ACTION_9 #undef ACTION_10 #undef ACTION_11 #undef ACTION_12 #undef ACTION_13 #undef ACTION_14 #undef ACTION_15 #undef ACTION_16 #undef CONDITION_B #undef CONDITION_C #undef CONDITION_D #undef CONDITION_E #undef CONDITION_G #undef CONDITION_H #undef CONDITION_I #undef CONDITION_J #undef CONDITION_K #undef CONDITION_M #undef CONDITION_N #undef CONDITION_O #undef CONDITION_P #undef CONDITION_R #undef CONDITION_S #undef CONDITION_T } // Second scan + analysis LabelT nLabels = flattenL(P, lunique); sop.init(nLabels); int r = 0; for (; r < e_rows; r += 2) { // Get rows pointer const PixelT * const img_row = img.ptr<PixelT>(r); const PixelT * const img_row_fol = (PixelT *)(((char*)img_row) + img.step.p[0]); LabelT * const img_labels_row = imgLabels.ptr<LabelT>(r); LabelT * const img_labels_row_fol = (LabelT *)(((char*)img_labels_row) + imgLabels.step.p[0]); int c = 0; for (; c < e_cols; c += 2) { LabelT iLabel = img_labels_row[c]; if (iLabel > 0) { iLabel = P[iLabel]; if (img_row[c] > 0) { img_labels_row[c] = iLabel; sop(r, c, iLabel); } else { img_labels_row[c] = 0; sop(r, c, 0); } if (img_row[c + 1] > 0) { img_labels_row[c + 1] = iLabel; sop(r, c + 1, iLabel); } else { img_labels_row[c + 1] = 0; sop(r, c + 1, 0); } if (img_row_fol[c] > 0) { img_labels_row_fol[c] = iLabel; sop(r + 1, c, iLabel); } else { img_labels_row_fol[c] = 0; sop(r + 1, c, 0); } if (img_row_fol[c + 1] > 0) { img_labels_row_fol[c + 1] = iLabel; sop(r + 1, c + 1, iLabel); } else { img_labels_row_fol[c + 1] = 0; sop(r + 1, c + 1, 0); } } else { img_labels_row[c] = 0; sop(r, c, 0); img_labels_row[c + 1] = 0; sop(r, c + 1, 0); img_labels_row_fol[c] = 0; sop(r + 1, c, 0); img_labels_row_fol[c + 1] = 0; sop(r + 1, c + 1, 0); } } // Last column if the number of columns is odd if (o_cols) { LabelT iLabel = img_labels_row[c]; if (iLabel > 0) { iLabel = P[iLabel]; if (img_row[c] > 0) { img_labels_row[c] = iLabel; sop(r, c, iLabel); } else { img_labels_row[c] = 0; sop(r, c, 0); } if (img_row_fol[c] > 0) { img_labels_row_fol[c] = iLabel; sop(r + 1, c, iLabel); } else { img_labels_row_fol[c] = 0; sop(r + 1, c, 0); } } else { img_labels_row[c] = 0; sop(r, c, 0); img_labels_row_fol[c] = 0; sop(r + 1, c, 0); } } } // Last row if the number of rows is odd if (o_rows) { // Get rows pointer const PixelT * const img_row = img.ptr<PixelT>(r); LabelT * const img_labels_row = imgLabels.ptr<LabelT>(r); int c = 0; for (; c < e_cols; c += 2) { LabelT iLabel = img_labels_row[c]; if (iLabel > 0) { iLabel = P[iLabel]; if (img_row[c] > 0) { img_labels_row[c] = iLabel; sop(r, c, iLabel); } else { img_labels_row[c] = 0; sop(r, c, 0); } if (img_row[c + 1] > 0) { img_labels_row[c + 1] = iLabel; sop(r, c + 1, iLabel); } else { img_labels_row[c + 1] = 0; sop(r, c + 1, 0); } } else { img_labels_row[c] = 0; sop(r, c, 0); img_labels_row[c + 1] = 0; sop(r, c + 1, 0); } } // Last column if the number of columns is odd if (o_cols) { LabelT iLabel = img_labels_row[c]; if (iLabel > 0) { iLabel = P[iLabel]; if (img_row[c] > 0) { img_labels_row[c] = iLabel; sop(r, c, iLabel); } else { img_labels_row[c] = 0; sop(r, c, 0); } } else { img_labels_row[c] = 0; sop(r, c, iLabel); } } } sop.finish(); return nLabels; }//End function LabelingBolelli operator() };//End struct LabelingBolelli //Parallel implementation of Spaghetti algorithm for 4-way connectivity, generated with the tool described in "One DAG to //Rule Them All", IEEE Transactions on Pattern Analysis and Machine Intelligence, Federico Bolelli et. al. //Parallelization method described in "Two More Strategies to Speed Up Connected Components Labeling Algorithms", //Image Analysis and Processing - ICIAP 2017, Federico Bolelli et. al. template<typename LabelT, typename PixelT, typename StatsOp = NoOp > struct LabelingBolelli4CParallel { class FirstScan : public cv::ParallelLoopBody { const cv::Mat& img_; cv::Mat& imgLabels_; LabelT* P_; int* chunksSizeAndLabels_; public: FirstScan(const cv::Mat& img, cv::Mat& imgLabels, LabelT* P, int* chunksSizeAndLabels) : img_(img), imgLabels_(imgLabels), P_(P), chunksSizeAndLabels_(chunksSizeAndLabels) {} FirstScan& operator=(const FirstScan&) { return *this; } void operator()(const cv::Range& range2) const CV_OVERRIDE { const Range range(range2.start * 2, std::min(range2.end * 2, img_.rows)); int r = range.start; chunksSizeAndLabels_[r] = range.end; LabelT label = stripeFirstLabel4Connectivity<LabelT>(r, imgLabels_.cols); const LabelT firstLabel = label; const int w = img_.cols; const int startR = r; { #define CONDITION_Q img_row_prev[c] > 0 #define CONDITION_S img_row[c - 1] > 0 #define CONDITION_X img_row[c] > 0 #define ACTION_1 img_labels_row[c] = 0; #define ACTION_2 img_labels_row[c] = label; \ P_[label] = label; \ label = label + 1; #define ACTION_3 img_labels_row[c] = img_labels_row_prev[c]; // x <- q #define ACTION_4 img_labels_row[c] = img_labels_row[c - 1]; // x <- s #define ACTION_5 img_labels_row[c] = set_union(P_, img_labels_row_prev[c], img_labels_row[c - 1]); // x <- q + s } // First row { const PixelT* const img_row = img_.ptr<PixelT>(r); LabelT* const img_labels_row = imgLabels_.ptr<LabelT>(r); int c = -1; goto fl_tree_0; fl_tree_0: if ((c += 1) >= w) goto fl_break; if (CONDITION_X) { ACTION_2 goto fl_tree_1; } else { ACTION_1 goto fl_tree_0; } fl_tree_1: if ((c += 1) >= w) goto fl_break; if (CONDITION_X) { ACTION_4 goto fl_tree_1; } else { ACTION_1 goto fl_tree_0; } fl_break:; } // Other rows ++r; for (; r < range.end; ++r) { // Get row pointers const PixelT* const img_row = img_.ptr<PixelT>(r); const PixelT* const img_row_prev = (PixelT*)(((char*)img_row) - img_.step.p[0]); LabelT* const img_labels_row = imgLabels_.ptr<LabelT>(r); LabelT* const img_labels_row_prev = (LabelT*)(((char*)img_labels_row) - imgLabels_.step.p[0]); int c = -1; goto cl_tree_0; cl_tree_0: if ((c += 1) >= w) goto cl_break; if (CONDITION_X) { if (CONDITION_Q) { ACTION_3 goto cl_tree_1; } else { ACTION_2 goto cl_tree_1; } } else { ACTION_1 goto cl_tree_0; } cl_tree_1: if ((c += 1) >= w) goto cl_break; if (CONDITION_X) { if (CONDITION_Q) { ACTION_5 goto cl_tree_1; } else { ACTION_4 goto cl_tree_1; } } else { ACTION_1 goto cl_tree_0; } cl_break:; } // undef conditions and actions { #undef ACTION_1 #undef ACTION_2 #undef ACTION_3 #undef ACTION_4 #undef ACTION_5 #undef CONDITION_Q #undef CONDITION_S #undef CONDITION_X } //write in the following memory location chunksSizeAndLabels_[startR + 1] = label - firstLabel; } }; class SecondScan : public cv::ParallelLoopBody { cv::Mat& imgLabels_; const LabelT* P_; StatsOp& sop_; StatsOp* sopArray_; LabelT& nLabels_; public: SecondScan(cv::Mat& imgLabels, const LabelT* P, StatsOp& sop, StatsOp* sopArray, LabelT& nLabels) : imgLabels_(imgLabels), P_(P), sop_(sop), sopArray_(sopArray), nLabels_(nLabels) {} SecondScan& operator=(const SecondScan&) { return *this; } void operator()(const cv::Range& range2) const CV_OVERRIDE { const Range range(range2.start * 2, std::min(range2.end * 2, imgLabels_.rows)); int r = range.start; const int rowBegin = r; const int rowEnd = range.end; if (rowBegin > 0) { sopArray_[rowBegin].initElement(nLabels_); sopArray_[rowBegin].setNextLoc(rowEnd); //_nextLoc = rowEnd; for (; r < rowEnd; ++r) { LabelT* img_row_start = imgLabels_.ptr<LabelT>(r); LabelT* const img_row_end = img_row_start + imgLabels_.cols; for (int c = 0; img_row_start != img_row_end; ++img_row_start, ++c) { *img_row_start = P_[*img_row_start]; sopArray_[rowBegin](r, c, *img_row_start); } } } else { //the first thread uses sop in order to make less merges sop_.setNextLoc(rowEnd); for (; r < rowEnd; ++r) { LabelT* img_row_start = imgLabels_.ptr<LabelT>(r); LabelT* const img_row_end = img_row_start + imgLabels_.cols; for (int c = 0; img_row_start != img_row_end; ++img_row_start, ++c) { *img_row_start = P_[*img_row_start]; sop_(r, c, *img_row_start); } } } } }; inline static void mergeLabels(cv::Mat& imgLabels, LabelT* P, const int* chunksSizeAndLabels) { // Merge Mask // +-+-+-+ // |-|q|-| // +-+-+-+ // |x| // +-+ const int w = imgLabels.cols, h = imgLabels.rows; for (int r = chunksSizeAndLabels[0]; r < h; r = chunksSizeAndLabels[r]) { LabelT* const imgLabels_row = imgLabels.ptr<LabelT>(r); LabelT* const imgLabels_row_prev = (LabelT*)(((char*)imgLabels_row) - imgLabels.step.p[0]); for (int c = 0; c < w; ++c) { #define condition_q imgLabels_row_prev[c] > 0 #define condition_x imgLabels_row[c] > 0 if (condition_x) { if (condition_q) { //merge of two label imgLabels_row[c] = set_union(P, imgLabels_row_prev[c], imgLabels_row[c]); } } } } #undef condition_q #undef condition_x } LabelT operator()(const cv::Mat& img, cv::Mat& imgLabels, int connectivity, StatsOp& sop) { CV_Assert(img.rows == imgLabels.rows); CV_Assert(img.cols == imgLabels.cols); CV_Assert(connectivity == 4); const int h = img.rows; const int w = img.cols; //A quick and dirty upper bound for the maximum number of labels. //Following formula comes from the fact that a 2x2 block in 4-way connectivity //labeling can never have more than 2 new labels and 1 label for background. //Worst case image example pattern: //1 0 1 0 1... //0 1 0 1 0... //1 0 1 0 1... //............ const size_t Plength = (size_t(h) * size_t(w) + 1) / 2 + 1; //Array used to store info and labeled pixel by each thread. //Different threads affect different memory location of chunksSizeAndLabels std::vector<int> chunksSizeAndLabels(roundUp(h, 2)); //Tree of labels std::vector<LabelT> P_(Plength, 0); LabelT* P = P_.data(); //First label is for background //P[0] = 0; cv::Range range2(0, divUp(h, 2)); const double nParallelStripes = std::max(1, std::min(h / 2, getNumThreads() * 4)); LabelT nLabels = 1; //First scan cv::parallel_for_(range2, FirstScan(img, imgLabels, P, chunksSizeAndLabels.data()), nParallelStripes); //merge labels of different chunks mergeLabels(imgLabels, P, chunksSizeAndLabels.data()); for (int i = 0; i < h; i = chunksSizeAndLabels[i]) { flattenL(P, stripeFirstLabel4Connectivity<int>(i, w), chunksSizeAndLabels[i + 1], nLabels); } //Array for statistics dataof threads std::vector<StatsOp> sopArray(h); sop.init(nLabels); //Second scan cv::parallel_for_(range2, SecondScan(imgLabels, P, sop, sopArray.data(), nLabels), nParallelStripes); StatsOp::mergeStats(imgLabels, sopArray.data(), sop, nLabels); sop.finish(); return nLabels; } };//End struct LabelingBolelli4CParallel //Implementation of Spaghetti algorithm for 4-way connectivity, generated with the tool described in "One DAG to //Rule Them All", IEEE Transactions on Pattern Analysis and Machine Intelligence, Federico Bolelli et. al. template<typename LabelT, typename PixelT, typename StatsOp = NoOp > struct LabelingBolelli4C { LabelT operator()(const cv::Mat& img, cv::Mat& imgLabels, int connectivity, StatsOp& sop) { CV_Assert(img.rows == imgLabels.rows); CV_Assert(img.cols == imgLabels.cols); CV_Assert(connectivity == 4); const int h = img.rows; const int w = img.cols; // A quick and dirty upper bound for the maximum number of labels. // Following formula comes from the fact that a 2x2 block in 4-connectivity case // can never have more than 2 new labels and 1 label for background. // Worst case image example pattern: // 1 0 1 0 1... // 0 1 0 1 0... // 1 0 1 0 1... // ............ const size_t Plength = size_t((size_t(h) * size_t(w) + 1) / 2) + 1; std::vector<LabelT> P_(Plength, 0); LabelT* P = P_.data(); P[0] = 0; LabelT lunique = 1; // First scan // We work with the 4-conn Rosenfeld mask // +-+ // |q| // +-+-+ // |s|x| // +-+-+ // A bunch of defines is used to check if the pixels are foreground // and to define actions to be performed { #define CONDITION_Q img_row_prev[c] > 0 #define CONDITION_S img_row[c - 1] > 0 #define CONDITION_X img_row[c] > 0 #define ACTION_1 img_labels_row[c] = 0; #define ACTION_2 img_labels_row[c] = lunique; \ P[lunique] = lunique; \ lunique = lunique + 1; // new label #define ACTION_3 img_labels_row[c] = img_labels_row_prev[c]; // x <- q #define ACTION_4 img_labels_row[c] = img_labels_row[c - 1]; // x <- s #define ACTION_5 img_labels_row[c] = set_union(P, img_labels_row_prev[c], img_labels_row[c - 1]); // x <- q + s } // First row { const PixelT* const img_row = img.ptr<PixelT>(0); LabelT* const img_labels_row = imgLabels.ptr<LabelT>(0); int c = -1; goto fl_tree_0; fl_tree_0: if ((c += 1) >= w) goto fl_break; if (CONDITION_X) { ACTION_2 goto fl_tree_1; } else { ACTION_1 goto fl_tree_0; } fl_tree_1: if ((c += 1) >= w) goto fl_break; if (CONDITION_X) { ACTION_4 goto fl_tree_1; } else { ACTION_1 goto fl_tree_0; } fl_break:; } // Other rows for (int r = 1; r < h; ++r) { // Get row pointers const PixelT* const img_row = img.ptr<PixelT>(r); const PixelT* const img_row_prev = (PixelT*)(((char*)img_row) - img.step.p[0]); LabelT* const img_labels_row = imgLabels.ptr<LabelT>(r); LabelT* const img_labels_row_prev = (LabelT*)(((char*)img_labels_row) - imgLabels.step.p[0]); int c = -1; goto cl_tree_0; cl_tree_0: if ((c += 1) >= w) goto cl_break; if (CONDITION_X) { if (CONDITION_Q) { ACTION_3 goto cl_tree_1; } else { ACTION_2 goto cl_tree_1; } } else { ACTION_1 goto cl_tree_0; } cl_tree_1: if ((c += 1) >= w) goto cl_break; if (CONDITION_X) { if (CONDITION_Q) { ACTION_5 goto cl_tree_1; } else { ACTION_4 goto cl_tree_1; } } else { ACTION_1 goto cl_tree_0; } cl_break:; } // undef conditions and actions { #undef ACTION_1 #undef ACTION_2 #undef ACTION_3 #undef ACTION_4 #undef ACTION_5 #undef CONDITION_Q #undef CONDITION_S #undef CONDITION_X } // Second scan + analysis LabelT nLabels = flattenL(P, lunique); sop.init(nLabels); for (int r = 0; r < h; ++r) { LabelT* img_row_start = imgLabels.ptr<LabelT>(r); LabelT* const img_row_end = img_row_start + w; for (int c = 0; img_row_start != img_row_end; ++img_row_start, ++c) { *img_row_start = P[*img_row_start]; sop(r, c, *img_row_start); } } sop.finish(); return nLabels; }//End function LabelingBolelli4C operator() };//End struct LabelingBolelli4C //Parallel implementation of Scan Array-based Union Find (SAUF) algorithm, as described in "Two More Strategies to Speed //Up Connected Components Labeling Algorithms", Image Analysis and Processing - ICIAP 2017, Federico Bolelli et. al. template<typename LabelT, typename PixelT, typename StatsOp = NoOp > struct LabelingWuParallel{ class FirstScan8Connectivity : public cv::ParallelLoopBody{ const cv::Mat& img_; cv::Mat& imgLabels_; LabelT *P_; int *chunksSizeAndLabels_; public: FirstScan8Connectivity(const cv::Mat& img, cv::Mat& imgLabels, LabelT *P, int *chunksSizeAndLabels) : img_(img), imgLabels_(imgLabels), P_(P), chunksSizeAndLabels_(chunksSizeAndLabels){} FirstScan8Connectivity& operator=(const FirstScan8Connectivity& ) { return *this; } void operator()(const cv::Range& range2) const CV_OVERRIDE { const Range range(range2.start * 2, std::min(range2.end * 2, img_.rows)); int r = range.start; chunksSizeAndLabels_[r] = range.end; LabelT label = stripeFirstLabel8Connectivity<LabelT>(r, imgLabels_.cols); const LabelT firstLabel = label; const int w = img_.cols; const int limitLine = r, startR = r; // Rosenfeld Mask // +-+-+-+ // |p|q|r| // +-+-+-+ // |s|x| // +-+-+ for (; r != range.end; ++r) { PixelT const * const img_row = img_.ptr<PixelT>(r); PixelT const * const img_row_prev = (PixelT *)(((char *)img_row) - img_.step.p[0]); LabelT * const imgLabels_row = imgLabels_.ptr<LabelT>(r); LabelT * const imgLabels_row_prev = (LabelT *)(((char *)imgLabels_row) - imgLabels_.step.p[0]); for (int c = 0; c < w; ++c) { #define condition_p c > 0 && r > limitLine && img_row_prev[c - 1] > 0 #define condition_q r > limitLine && img_row_prev[c] > 0 #define condition_r c < w - 1 && r > limitLine && img_row_prev[c + 1] > 0 #define condition_s c > 0 && img_row[c - 1] > 0 #define condition_x img_row[c] > 0 if (condition_x){ if (condition_q){ //copy q imgLabels_row[c] = imgLabels_row_prev[c]; } else{ //not q if (condition_r){ if (condition_p){ //concavity p->x->r. Merge imgLabels_row[c] = set_union(P_, imgLabels_row_prev[c - 1], imgLabels_row_prev[c + 1]); } else{ //not p and q if (condition_s){ //step s->x->r. Merge imgLabels_row[c] = set_union(P_, imgLabels_row[c - 1], imgLabels_row_prev[c + 1]); } else{ //not p, q and s //copy r imgLabels_row[c] = imgLabels_row_prev[c + 1]; } } } else{ //not r and q if (condition_p){ //copy p imgLabels_row[c] = imgLabels_row_prev[c - 1]; } else{//not r,q and p if (condition_s){ imgLabels_row[c] = imgLabels_row[c - 1]; } else{ //new label imgLabels_row[c] = label; P_[label] = label; label = label + 1; } } } } } else{ //x is a background pixel imgLabels_row[c] = 0; } } } //write in the follower memory location chunksSizeAndLabels_[startR + 1] = label - firstLabel; } #undef condition_p #undef condition_q #undef condition_r #undef condition_s #undef condition_x }; class FirstScan4Connectivity : public cv::ParallelLoopBody{ const cv::Mat& img_; cv::Mat& imgLabels_; LabelT *P_; int *chunksSizeAndLabels_; public: FirstScan4Connectivity(const cv::Mat& img, cv::Mat& imgLabels, LabelT *P, int *chunksSizeAndLabels) : img_(img), imgLabels_(imgLabels), P_(P), chunksSizeAndLabels_(chunksSizeAndLabels){} FirstScan4Connectivity& operator=(const FirstScan4Connectivity& ) { return *this; } void operator()(const cv::Range& range2) const CV_OVERRIDE { const Range range(range2.start * 2, std::min(range2.end * 2, img_.rows)); int r = range.start; chunksSizeAndLabels_[r] = range.end; LabelT label = stripeFirstLabel4Connectivity<LabelT>(r, imgLabels_.cols); const LabelT firstLabel = label; const int w = img_.cols; const int limitLine = r, startR = r; // Rosenfeld Mask // +-+-+-+ // |-|q|-| // +-+-+-+ // |s|x| // +-+-+ for (; r != range.end; ++r){ PixelT const * const img_row = img_.ptr<PixelT>(r); PixelT const * const img_row_prev = (PixelT *)(((char *)img_row) - img_.step.p[0]); LabelT * const imgLabels_row = imgLabels_.ptr<LabelT>(r); LabelT * const imgLabels_row_prev = (LabelT *)(((char *)imgLabels_row) - imgLabels_.step.p[0]); for (int c = 0; c < w; ++c) { #define condition_q r > limitLine && img_row_prev[c] > 0 #define condition_s c > 0 && img_row[c - 1] > 0 #define condition_x img_row[c] > 0 if (condition_x){ if (condition_q){ if (condition_s){ //step s->x->q. Merge imgLabels_row[c] = set_union(P_, imgLabels_row[c - 1], imgLabels_row_prev[c]); } else{ //copy q imgLabels_row[c] = imgLabels_row_prev[c]; } } else{ if (condition_s){ // copy s imgLabels_row[c] = imgLabels_row[c - 1]; } else{ //new label imgLabels_row[c] = label; P_[label] = label; label = label + 1; } } } else{ //x is a background pixel imgLabels_row[c] = 0; } } } //write in the following memory location chunksSizeAndLabels_[startR + 1] = label - firstLabel; } #undef condition_q #undef condition_s #undef condition_x }; class SecondScan : public cv::ParallelLoopBody{ cv::Mat& imgLabels_; const LabelT *P_; StatsOp& sop_; StatsOp *sopArray_; LabelT& nLabels_; public: SecondScan(cv::Mat& imgLabels, const LabelT *P, StatsOp& sop, StatsOp *sopArray, LabelT& nLabels) : imgLabels_(imgLabels), P_(P), sop_(sop), sopArray_(sopArray), nLabels_(nLabels){} SecondScan& operator=(const SecondScan& ) { return *this; } void operator()(const cv::Range& range2) const CV_OVERRIDE { const Range range(range2.start * 2, std::min(range2.end * 2, imgLabels_.rows)); int r = range.start; const int rowBegin = r; const int rowEnd = range.end; if (rowBegin > 0){ sopArray_[rowBegin].initElement(nLabels_); sopArray_[rowBegin].setNextLoc(rowEnd); //_nextLoc = rowEnd; for (; r < rowEnd; ++r) { LabelT * img_row_start = imgLabels_.ptr<LabelT>(r); LabelT * const img_row_end = img_row_start + imgLabels_.cols; for (int c = 0; img_row_start != img_row_end; ++img_row_start, ++c){ *img_row_start = P_[*img_row_start]; sopArray_[rowBegin](r, c, *img_row_start); } } } else{ //the first thread uses sop in order to make less merges sop_.setNextLoc(rowEnd); for (; r < rowEnd; ++r) { LabelT * img_row_start = imgLabels_.ptr<LabelT>(r); LabelT * const img_row_end = img_row_start + imgLabels_.cols; for (int c = 0; img_row_start != img_row_end; ++img_row_start, ++c){ *img_row_start = P_[*img_row_start]; sop_(r, c, *img_row_start); } } } } }; inline static void mergeLabels8Connectivity(cv::Mat& imgLabels, LabelT *P, const int *chunksSizeAndLabels){ // Merge Mask // +-+-+-+ // |p|q|r| // +-+-+-+ // |x| // +-+ const int w = imgLabels.cols, h = imgLabels.rows; for (int r = chunksSizeAndLabels[0]; r < h; r = chunksSizeAndLabels[r]){ LabelT * const imgLabels_row = imgLabels.ptr<LabelT>(r); LabelT * const imgLabels_row_prev = (LabelT *)(((char *)imgLabels_row) - imgLabels.step.p[0]); for (int c = 0; c < w; ++c){ #define condition_p c > 0 && imgLabels_row_prev[c - 1] > 0 #define condition_q imgLabels_row_prev[c] > 0 #define condition_r c < w - 1 && imgLabels_row_prev[c + 1] > 0 #define condition_x imgLabels_row[c] > 0 if (condition_x){ if (condition_p){ //merge of two label imgLabels_row[c] = set_union(P, imgLabels_row_prev[c - 1], imgLabels_row[c]); } if (condition_r){ //merge of two label imgLabels_row[c] = set_union(P, imgLabels_row_prev[c + 1], imgLabels_row[c]); } if (condition_q){ //merge of two label imgLabels_row[c] = set_union(P, imgLabels_row_prev[c], imgLabels_row[c]); } } } } #undef condition_p #undef condition_q #undef condition_r #undef condition_x } inline static void mergeLabels4Connectivity(cv::Mat& imgLabels, LabelT *P, const int *chunksSizeAndLabels){ // Merge Mask // +-+-+-+ // |-|q|-| // +-+-+-+ // |x| // +-+ const int w = imgLabels.cols, h = imgLabels.rows; for (int r = chunksSizeAndLabels[0]; r < h; r = chunksSizeAndLabels[r]){ LabelT * const imgLabels_row = imgLabels.ptr<LabelT>(r); LabelT * const imgLabels_row_prev = (LabelT *)(((char *)imgLabels_row) - imgLabels.step.p[0]); for (int c = 0; c < w; ++c){ #define condition_q imgLabels_row_prev[c] > 0 #define condition_x imgLabels_row[c] > 0 if (condition_x){ if (condition_q){ //merge of two label imgLabels_row[c] = set_union(P, imgLabels_row_prev[c], imgLabels_row[c]); } } } } #undef condition_q #undef condition_x } LabelT operator()(const cv::Mat& img, cv::Mat& imgLabels, int connectivity, StatsOp& sop){ CV_Assert(img.rows == imgLabels.rows); CV_Assert(img.cols == imgLabels.cols); CV_Assert(connectivity == 8 || connectivity == 4); const int h = img.rows; const int w = img.cols; //A quick and dirty upper bound for the maximum number of labels. //Following formula comes from the fact that a 2x2 block in 4-way connectivity //labeling can never have more than 2 new labels and 1 label for background. //Worst case image example pattern: //1 0 1 0 1... //0 1 0 1 0... //1 0 1 0 1... //............ //Obviously, 4-way connectivity upper bound is also good for 8-way connectivity labeling const size_t Plength = (size_t(h) * size_t(w) + 1) / 2 + 1; //Array used to store info and labeled pixel by each thread. //Different threads affect different memory location of chunksSizeAndLabels std::vector<int> chunksSizeAndLabels(roundUp(h, 2)); //Tree of labels std::vector<LabelT> P_(Plength, 0); LabelT *P = P_.data(); //First label is for background //P[0] = 0; cv::Range range2(0, divUp(h, 2)); const double nParallelStripes = std::max(1, std::min(h / 2, getNumThreads()*4)); LabelT nLabels = 1; if (connectivity == 8){ //First scan cv::parallel_for_(range2, FirstScan8Connectivity(img, imgLabels, P, chunksSizeAndLabels.data()), nParallelStripes); //merge labels of different chunks mergeLabels8Connectivity(imgLabels, P, chunksSizeAndLabels.data()); for (int i = 0; i < h; i = chunksSizeAndLabels[i]){ flattenL(P, stripeFirstLabel8Connectivity<int>(i, w), chunksSizeAndLabels[i + 1], nLabels); } } else{ //First scan cv::parallel_for_(range2, FirstScan4Connectivity(img, imgLabels, P, chunksSizeAndLabels.data()), nParallelStripes); //merge labels of different chunks mergeLabels4Connectivity(imgLabels, P, chunksSizeAndLabels.data()); for (int i = 0; i < h; i = chunksSizeAndLabels[i]){ flattenL(P, stripeFirstLabel4Connectivity<int>(i, w), chunksSizeAndLabels[i + 1], nLabels); } } //Array for statistics dataof threads std::vector<StatsOp> sopArray(h); sop.init(nLabels); //Second scan cv::parallel_for_(range2, SecondScan(imgLabels, P, sop, sopArray.data(), nLabels), nParallelStripes); StatsOp::mergeStats(imgLabels, sopArray.data(), sop, nLabels); sop.finish(); return nLabels; } };//End struct LabelingWuParallel //Based on "Two Strategies to Speed up Connected Components Algorithms", the SAUF (Scan Array-based Union Find) variant //using decision trees, Kesheng Wu et. al. template<typename LabelT, typename PixelT, typename StatsOp = NoOp > struct LabelingWu{ LabelT operator()(const cv::Mat& img, cv::Mat& imgLabels, int connectivity, StatsOp& sop){ CV_Assert(imgLabels.rows == img.rows); CV_Assert(imgLabels.cols == img.cols); CV_Assert(connectivity == 8 || connectivity == 4); const int h = img.rows; const int w = img.cols; //A quick and dirty upper bound for the maximum number of labels. //Following formula comes from the fact that a 2x2 block in 4-way connectivity //labeling can never have more than 2 new labels and 1 label for background. //Worst case image example pattern: //1 0 1 0 1... //0 1 0 1 0... //1 0 1 0 1... //............ //Obviously, 4-way connectivity upper bound is also good for 8-way connectivity labeling const size_t Plength = (size_t(h) * size_t(w) + 1) / 2 + 1; //array P for equivalences resolution std::vector<LabelT> P_(Plength, 0); LabelT *P = P_.data(); //first label is for background pixels //P[0] = 0; LabelT lunique = 1; if (connectivity == 8){ for (int r = 0; r < h; ++r){ // Get row pointers PixelT const * const img_row = img.ptr<PixelT>(r); PixelT const * const img_row_prev = (PixelT *)(((char *)img_row) - img.step.p[0]); LabelT * const imgLabels_row = imgLabels.ptr<LabelT>(r); LabelT * const imgLabels_row_prev = (LabelT *)(((char *)imgLabels_row) - imgLabels.step.p[0]); for (int c = 0; c < w; ++c){ #define condition_p c>0 && r>0 && img_row_prev[c - 1]>0 #define condition_q r>0 && img_row_prev[c]>0 #define condition_r c < w - 1 && r > 0 && img_row_prev[c + 1] > 0 #define condition_s c > 0 && img_row[c - 1] > 0 #define condition_x img_row[c] > 0 if (condition_x){ if (condition_q){ //x <- q imgLabels_row[c] = imgLabels_row_prev[c]; } else{ // q = 0 if (condition_r){ if (condition_p){ // x <- merge(p,r) imgLabels_row[c] = set_union(P, imgLabels_row_prev[c - 1], imgLabels_row_prev[c + 1]); } else{ // p = q = 0 if (condition_s){ // x <- merge(s,r) imgLabels_row[c] = set_union(P, imgLabels_row[c - 1], imgLabels_row_prev[c + 1]); } else{ // p = q = s = 0 // x <- r imgLabels_row[c] = imgLabels_row_prev[c + 1]; } } } else{ // r = q = 0 if (condition_p){ // x <- p imgLabels_row[c] = imgLabels_row_prev[c - 1]; } else{ // r = q = p = 0 if (condition_s){ imgLabels_row[c] = imgLabels_row[c - 1]; } else{ //new label imgLabels_row[c] = lunique; P[lunique] = lunique; lunique = lunique + 1; } } } } } else{ //x is a background pixel imgLabels_row[c] = 0; } } } #undef condition_p #undef condition_q #undef condition_r #undef condition_s #undef condition_x } else{ for (int r = 0; r < h; ++r){ PixelT const * const img_row = img.ptr<PixelT>(r); PixelT const * const img_row_prev = (PixelT *)(((char *)img_row) - img.step.p[0]); LabelT * const imgLabels_row = imgLabels.ptr<LabelT>(r); LabelT * const imgLabels_row_prev = (LabelT *)(((char *)imgLabels_row) - imgLabels.step.p[0]); for (int c = 0; c < w; ++c) { #define condition_q r > 0 && img_row_prev[c] > 0 #define condition_s c > 0 && img_row[c - 1] > 0 #define condition_x img_row[c] > 0 if (condition_x){ if (condition_q){ if (condition_s){ //Merge s->x->q imgLabels_row[c] = set_union(P, imgLabels_row[c - 1], imgLabels_row_prev[c]); } else{ //copy q imgLabels_row[c] = imgLabels_row_prev[c]; } } else{ if (condition_s){ // copy s imgLabels_row[c] = imgLabels_row[c - 1]; } else{ //new label imgLabels_row[c] = lunique; P[lunique] = lunique; lunique = lunique + 1; } } } else{ //x is a background pixel imgLabels_row[c] = 0; } } } #undef condition_q #undef condition_s #undef condition_x } //analysis LabelT nLabels = flattenL(P, lunique); sop.init(nLabels); for (int r = 0; r < h; ++r) { LabelT * img_row_start = imgLabels.ptr<LabelT>(r); LabelT * const img_row_end = img_row_start + w; for (int c = 0; img_row_start != img_row_end; ++img_row_start, ++c){ *img_row_start = P[*img_row_start]; sop(r, c, *img_row_start); } } sop.finish(); return nLabels; }//End function LabelingWu operator() };//End struct LabelingWu //Parallel implementation of BBDT (Block-Based with Decision Tree) algorithm, as described in "Two More Strategies to Speed //Up Connected Components Labeling Algorithms", Image Analysis and Processing - ICIAP 2017, Federico Bolelli et. al. template<typename LabelT, typename PixelT, typename StatsOp = NoOp > struct LabelingGranaParallel{ class FirstScan : public cv::ParallelLoopBody{ private: const cv::Mat& img_; cv::Mat& imgLabels_; LabelT *P_; int *chunksSizeAndLabels_; public: FirstScan(const cv::Mat& img, cv::Mat& imgLabels, LabelT *P, int *chunksSizeAndLabels) : img_(img), imgLabels_(imgLabels), P_(P), chunksSizeAndLabels_(chunksSizeAndLabels){} FirstScan& operator=(const FirstScan&) { return *this; } void operator()(const cv::Range& range2) const CV_OVERRIDE { const Range range(range2.start * 2, std::min(range2.end * 2, img_.rows)); int r = range.start; chunksSizeAndLabels_[r] = range.end; LabelT label = stripeFirstLabel8Connectivity<LabelT>(r, imgLabels_.cols); const LabelT firstLabel = label; const int h = img_.rows, w = img_.cols; const int limitLine = r + 1, startR = r; for (; r < range.end; r += 2){ // Get rows pointer const PixelT * const img_row = img_.ptr<uchar>(r); const PixelT * const img_row_prev = (PixelT *)(((char *)img_row) - img_.step.p[0]); const PixelT * const img_row_prev_prev = (PixelT *)(((char *)img_row_prev) - img_.step.p[0]); const PixelT * const img_row_fol = (PixelT *)(((char *)img_row) + img_.step.p[0]); LabelT * const imgLabels_row = imgLabels_.ptr<LabelT>(r); LabelT * const imgLabels_row_prev_prev = (LabelT *)(((char *)imgLabels_row) - imgLabels_.step.p[0] - imgLabels_.step.p[0]); for (int c = 0; c < w; c += 2) { // We work with 2x2 blocks // +-+-+-+ // |P|Q|R| // +-+-+-+ // |S|X| // +-+-+ // The pixels are named as follows // +---+---+---+ // |a b|c d|e f| // |g h|i j|k l| // +---+---+---+ // |m n|o p| // |q r|s t| // +---+---+ // Pixels a, f, l, q are not needed, since we need to understand the // the connectivity between these blocks and those pixels only matter // when considering the outer connectivities // A bunch of defines used to check if the pixels are foreground, // without going outside the image limits. #define condition_b c-1>=0 && r > limitLine && img_row_prev_prev[c-1]>0 #define condition_c r > limitLine && img_row_prev_prev[c]>0 #define condition_d c+1<w && r > limitLine && img_row_prev_prev[c+1]>0 #define condition_e c+2<w && r > limitLine && img_row_prev_prev[c+2]>0 #define condition_g c-2>=0 && r > limitLine - 1 && img_row_prev[c-2]>0 #define condition_h c-1>=0 && r > limitLine - 1 && img_row_prev[c-1]>0 #define condition_i r > limitLine - 1 && img_row_prev[c]>0 #define condition_j c+1<w && r > limitLine - 1 && img_row_prev[c+1]>0 #define condition_k c+2<w && r > limitLine - 1 && img_row_prev[c+2]>0 #define condition_m c-2>=0 && img_row[c-2]>0 #define condition_n c-1>=0 && img_row[c-1]>0 #define condition_o img_row[c]>0 #define condition_p c+1<w && img_row[c+1]>0 #define condition_r c-1>=0 && r+1<h && img_row_fol[c-1]>0 #define condition_s r+1<h && img_row_fol[c]>0 #define condition_t c+1<w && r+1<h && img_row_fol[c+1]>0 // This is a decision tree which allows to choose which action to // perform, checking as few conditions as possible. // Actions are available after the tree. if (condition_o) { if (condition_n) { if (condition_j) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { if (condition_h) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_g) { if (condition_b) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } else { if (condition_p) { if (condition_k) { if (condition_d) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { if (condition_h) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_g) { if (condition_b) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } else { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } } else { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } } } else { if (condition_r) { if (condition_j) { if (condition_m) { if (condition_h) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } else { if (condition_g) { if (condition_b) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } else { if (condition_i) { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } else { if (condition_h) { if (condition_c) { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } else { //Action_14: Merge labels of block P_, Q and S imgLabels_row[c] = set_union(P_, set_union(P_, imgLabels_row_prev_prev[c - 2], imgLabels_row_prev_prev[c]), imgLabels_row[c - 2]); continue; } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } } else { if (condition_p) { if (condition_k) { if (condition_m) { if (condition_h) { if (condition_d) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } else { if (condition_d) { if (condition_g) { if (condition_b) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } else { if (condition_i) { if (condition_g) { if (condition_b) { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } else { //Action_16: labels of block Q, R and S imgLabels_row[c] = set_union(P_, set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row_prev_prev[c + 2]), imgLabels_row[c - 2]); continue; } } else { //Action_16: labels of block Q, R and S imgLabels_row[c] = set_union(P_, set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row_prev_prev[c + 2]), imgLabels_row[c - 2]); continue; } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } } else { if (condition_i) { if (condition_d) { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } else { //Action_16: labels of block Q, R and S imgLabels_row[c] = set_union(P_, set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row_prev_prev[c + 2]), imgLabels_row[c - 2]); continue; } } else { if (condition_h) { if (condition_d) { if (condition_c) { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } else { //Action_15: Merge labels of block P_, R and S imgLabels_row[c] = set_union(P_, set_union(P_, imgLabels_row_prev_prev[c - 2], imgLabels_row_prev_prev[c + 2]), imgLabels_row[c - 2]); continue; } } else { //Action_15: Merge labels of block P_, R and S imgLabels_row[c] = set_union(P_, set_union(P_, imgLabels_row_prev_prev[c - 2], imgLabels_row_prev_prev[c + 2]), imgLabels_row[c - 2]); continue; } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } } else { if (condition_h) { if (condition_m) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { // ACTION_9 Merge labels of block P_ and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c - 2], imgLabels_row[c - 2]); continue; } } else { if (condition_i) { if (condition_m) { if (condition_g) { if (condition_b) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } } } } else { if (condition_h) { if (condition_m) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { // ACTION_9 Merge labels of block P_ and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c - 2], imgLabels_row[c - 2]); continue; } } else { if (condition_i) { if (condition_m) { if (condition_g) { if (condition_b) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } } } } } else { if (condition_j) { if (condition_i) { //Action_4: Assign label of block Q imgLabels_row[c] = imgLabels_row_prev_prev[c]; continue; } else { if (condition_h) { if (condition_c) { //Action_4: Assign label of block Q imgLabels_row[c] = imgLabels_row_prev_prev[c]; continue; } else { //Action_7: Merge labels of block P_ and Q imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c - 2], imgLabels_row_prev_prev[c]); continue; } } else { //Action_4: Assign label of block Q imgLabels_row[c] = imgLabels_row_prev_prev[c]; continue; } } } else { if (condition_p) { if (condition_k) { if (condition_i) { if (condition_d) { //Action_5: Assign label of block R imgLabels_row[c] = imgLabels_row_prev_prev[c + 2]; continue; } else { // ACTION_10 Merge labels of block Q and R imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row_prev_prev[c + 2]); continue; } } else { if (condition_h) { if (condition_d) { if (condition_c) { //Action_5: Assign label of block R imgLabels_row[c] = imgLabels_row_prev_prev[c + 2]; continue; } else { //Action_8: Merge labels of block P_ and R imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c - 2], imgLabels_row_prev_prev[c + 2]); continue; } } else { //Action_8: Merge labels of block P_ and R imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c - 2], imgLabels_row_prev_prev[c + 2]); continue; } } else { //Action_5: Assign label of block R imgLabels_row[c] = imgLabels_row_prev_prev[c + 2]; continue; } } } else { if (condition_i) { //Action_4: Assign label of block Q imgLabels_row[c] = imgLabels_row_prev_prev[c]; continue; } else { if (condition_h) { //Action_3: Assign label of block P_ imgLabels_row[c] = imgLabels_row_prev_prev[c - 2]; continue; } else { //Action_2: New label (the block has foreground pixels and is not connected to anything else) imgLabels_row[c] = label; P_[label] = label; label = label + 1; continue; } } } } else { if (condition_i) { //Action_4: Assign label of block Q imgLabels_row[c] = imgLabels_row_prev_prev[c]; continue; } else { if (condition_h) { //Action_3: Assign label of block P_ imgLabels_row[c] = imgLabels_row_prev_prev[c - 2]; continue; } else { //Action_2: New label (the block has foreground pixels and is not connected to anything else) imgLabels_row[c] = label; P_[label] = label; label = label + 1; continue; } } } } } } } else { if (condition_s) { if (condition_p) { if (condition_n) { if (condition_j) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { if (condition_h) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_g) { if (condition_b) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } else { if (condition_k) { if (condition_d) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { if (condition_h) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_g) { if (condition_b) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } else { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } } } else { if (condition_r) { if (condition_j) { if (condition_m) { if (condition_h) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } else { if (condition_g) { if (condition_b) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { if (condition_k) { if (condition_d) { if (condition_m) { if (condition_h) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } else { if (condition_g) { if (condition_b) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } else { if (condition_i) { if (condition_m) { if (condition_h) { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } else { if (condition_g) { if (condition_b) { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } else { //Action_16: labels of block Q, R and S imgLabels_row[c] = set_union(P_, set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row_prev_prev[c + 2]), imgLabels_row[c - 2]); continue; } } else { //Action_16: labels of block Q, R and S imgLabels_row[c] = set_union(P_, set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row_prev_prev[c + 2]), imgLabels_row[c - 2]); continue; } } } else { //Action_16: labels of block Q, R and S imgLabels_row[c] = set_union(P_, set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row_prev_prev[c + 2]), imgLabels_row[c - 2]); continue; } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } else { if (condition_i) { if (condition_m) { if (condition_h) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_g) { if (condition_b) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } } } } else { if (condition_j) { //Action_4: Assign label of block Q imgLabels_row[c] = imgLabels_row_prev_prev[c]; continue; } else { if (condition_k) { if (condition_i) { if (condition_d) { //Action_5: Assign label of block R imgLabels_row[c] = imgLabels_row_prev_prev[c + 2]; continue; } else { // ACTION_10 Merge labels of block Q and R imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row_prev_prev[c + 2]); continue; } } else { //Action_5: Assign label of block R imgLabels_row[c] = imgLabels_row_prev_prev[c + 2]; continue; } } else { if (condition_i) { //Action_4: Assign label of block Q imgLabels_row[c] = imgLabels_row_prev_prev[c]; continue; } else { //Action_2: New label (the block has foreground pixels and is not connected to anything else) imgLabels_row[c] = label; P_[label] = label; label = label + 1; continue; } } } } } } else { if (condition_r) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_n) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_2: New label (the block has foreground pixels and is not connected to anything else) imgLabels_row[c] = label; P_[label] = label; label = label + 1; continue; } } } } else { if (condition_p) { if (condition_j) { //Action_4: Assign label of block Q imgLabels_row[c] = imgLabels_row_prev_prev[c]; continue; } else { if (condition_k) { if (condition_i) { if (condition_d) { //Action_5: Assign label of block R imgLabels_row[c] = imgLabels_row_prev_prev[c + 2]; continue; } else { // ACTION_10 Merge labels of block Q and R imgLabels_row[c] = set_union(P_, imgLabels_row_prev_prev[c], imgLabels_row_prev_prev[c + 2]); continue; } } else { //Action_5: Assign label of block R imgLabels_row[c] = imgLabels_row_prev_prev[c + 2]; continue; } } else { if (condition_i) { //Action_4: Assign label of block Q imgLabels_row[c] = imgLabels_row_prev_prev[c]; continue; } else { //Action_2: New label (the block has foreground pixels and is not connected to anything else) imgLabels_row[c] = label; P_[label] = label; label = label + 1; continue; } } } } else { if (condition_t) { //Action_2: New label (the block has foreground pixels and is not connected to anything else) imgLabels_row[c] = label; P_[label] = label; label = label + 1; continue; } else { // Action_1: No action (the block has no foreground pixels) imgLabels_row[c] = 0; continue; } } } } } } //write in the follower memory location chunksSizeAndLabels_[startR + 1] = label - firstLabel; } #undef condition_k #undef condition_j #undef condition_i #undef condition_h #undef condition_g #undef condition_e #undef condition_d #undef condition_c #undef condition_b }; class SecondScan : public cv::ParallelLoopBody{ private: const cv::Mat& img_; cv::Mat& imgLabels_; LabelT *P_; StatsOp& sop_; StatsOp *sopArray_; LabelT& nLabels_; public: SecondScan(const cv::Mat& img, cv::Mat& imgLabels, LabelT *P, StatsOp& sop, StatsOp *sopArray, LabelT& nLabels) : img_(img), imgLabels_(imgLabels), P_(P), sop_(sop), sopArray_(sopArray), nLabels_(nLabels){} void operator()(const cv::Range& range2) const CV_OVERRIDE { const Range range(range2.start * 2, std::min(range2.end * 2, img_.rows)); int r = range.start; const int rowBegin = r; const int rowEnd = range.end; if (rowBegin > 0){ sopArray_[rowBegin].initElement(nLabels_); sopArray_[rowBegin].setNextLoc(rowEnd); //_nextLoc = rowEnd; if (imgLabels_.rows& 1){ if (imgLabels_.cols& 1){ //Case 1: both rows and cols odd for (; r < rowEnd; r += 2){ // Get rows pointer const PixelT * const img_row = img_.ptr<PixelT>(r); const PixelT * const img_row_fol = (PixelT *)(((char *)img_row) + img_.step.p[0]); LabelT * const imgLabels_row = imgLabels_.ptr<LabelT>(r); LabelT * const imgLabels_row_fol = (LabelT *)(((char *)imgLabels_row) + imgLabels_.step.p[0]); // Get rows pointer for (int c = 0; c < imgLabels_.cols; c += 2) { LabelT iLabel = imgLabels_row[c]; if (iLabel > 0) { iLabel = P_[iLabel]; if (img_row[c] > 0){ imgLabels_row[c] = iLabel; sopArray_[rowBegin](r, c, iLabel); } else{ imgLabels_row[c] = 0; sopArray_[rowBegin](r, c, 0); } if (c + 1 < imgLabels_.cols) { if (img_row[c + 1] > 0){ imgLabels_row[c + 1] = iLabel; sopArray_[rowBegin](r, c + 1, iLabel); } else{ imgLabels_row[c + 1] = 0; sopArray_[rowBegin](r, c + 1, 0); } if (r + 1 < imgLabels_.rows) { if (img_row_fol[c] > 0){ imgLabels_row_fol[c] = iLabel; sopArray_[rowBegin](r + 1, c, iLabel); } else{ imgLabels_row_fol[c] = 0; sopArray_[rowBegin](r + 1, c, 0); } if (img_row_fol[c + 1] > 0){ imgLabels_row_fol[c + 1] = iLabel; sopArray_[rowBegin](r + 1, c + 1, iLabel); } else{ imgLabels_row_fol[c + 1] = 0; sopArray_[rowBegin](r + 1, c + 1, 0); } } } else if (r + 1 < imgLabels_.rows) { if (img_row_fol[c] > 0){ imgLabels_row_fol[c] = iLabel; sopArray_[rowBegin](r + 1, c, iLabel); } else{ imgLabels_row_fol[c] = 0; sopArray_[rowBegin](r + 1, c, 0); } } } else { imgLabels_row[c] = 0; sopArray_[rowBegin](r, c, 0); if (c + 1 < imgLabels_.cols) { imgLabels_row[c + 1] = 0; sopArray_[rowBegin](r, c + 1, 0); if (r + 1 < imgLabels_.rows) { imgLabels_row_fol[c] = 0; imgLabels_row_fol[c + 1] = 0; sopArray_[rowBegin](r + 1, c, 0); sopArray_[rowBegin](r + 1, c + 1, 0); } } else if (r + 1 < imgLabels_.rows) { imgLabels_row_fol[c] = 0; sopArray_[rowBegin](r + 1, c, 0); } } } } }//END Case 1 else{ //Case 2: only rows odd for (; r < rowEnd; r += 2){ // Get rows pointer const PixelT * const img_row = img_.ptr<PixelT>(r); const PixelT * const img_row_fol = (PixelT *)(((char *)img_row) + img_.step.p[0]); LabelT * const imgLabels_row = imgLabels_.ptr<LabelT>(r); LabelT * const imgLabels_row_fol = (LabelT *)(((char *)imgLabels_row) + imgLabels_.step.p[0]); // Get rows pointer for (int c = 0; c < imgLabels_.cols; c += 2) { LabelT iLabel = imgLabels_row[c]; if (iLabel > 0) { iLabel = P_[iLabel]; if (img_row[c] > 0){ imgLabels_row[c] = iLabel; sopArray_[rowBegin](r, c, iLabel); } else{ imgLabels_row[c] = 0; sopArray_[rowBegin](r, c, 0); } if (img_row[c + 1] > 0){ imgLabels_row[c + 1] = iLabel; sopArray_[rowBegin](r, c + 1, iLabel); } else{ imgLabels_row[c + 1] = 0; sopArray_[rowBegin](r, c + 1, 0); } if (r + 1 < imgLabels_.rows) { if (img_row_fol[c] > 0){ imgLabels_row_fol[c] = iLabel; sopArray_[rowBegin](r + 1, c, iLabel); } else{ imgLabels_row_fol[c] = 0; sopArray_[rowBegin](r + 1, c, 0); } if (img_row_fol[c + 1] > 0){ imgLabels_row_fol[c + 1] = iLabel; sopArray_[rowBegin](r + 1, c + 1, iLabel); } else{ imgLabels_row_fol[c + 1] = 0; sopArray_[rowBegin](r + 1, c + 1, 0); } } } else { imgLabels_row[c] = 0; imgLabels_row[c + 1] = 0; sopArray_[rowBegin](r, c, 0); sopArray_[rowBegin](r, c + 1, 0); if (r + 1 < imgLabels_.rows) { imgLabels_row_fol[c] = 0; imgLabels_row_fol[c + 1] = 0; sopArray_[rowBegin](r + 1, c, 0); sopArray_[rowBegin](r + 1, c + 1, 0); } } } } }// END Case 2 } else{ if (imgLabels_.cols& 1){ //Case 3: only cols odd for (; r < rowEnd; r += 2){ // Get rows pointer const PixelT * const img_row = img_.ptr<PixelT>(r); const PixelT * const img_row_fol = (PixelT *)(((char *)img_row) + img_.step.p[0]); LabelT * const imgLabels_row = imgLabels_.ptr<LabelT>(r); LabelT * const imgLabels_row_fol = (LabelT *)(((char *)imgLabels_row) + imgLabels_.step.p[0]); // Get rows pointer for (int c = 0; c < imgLabels_.cols; c += 2) { LabelT iLabel = imgLabels_row[c]; if (iLabel > 0) { iLabel = P_[iLabel]; if (img_row[c] > 0){ imgLabels_row[c] = iLabel; sopArray_[rowBegin](r, c, iLabel); } else{ imgLabels_row[c] = 0; sopArray_[rowBegin](r, c, 0); } if (img_row_fol[c] > 0){ imgLabels_row_fol[c] = iLabel; sopArray_[rowBegin](r + 1, c, iLabel); } else{ imgLabels_row_fol[c] = 0; sopArray_[rowBegin](r + 1, c, 0); } if (c + 1 < imgLabels_.cols) { if (img_row[c + 1] > 0){ imgLabels_row[c + 1] = iLabel; sopArray_[rowBegin](r, c + 1, iLabel); } else{ imgLabels_row[c + 1] = 0; sopArray_[rowBegin](r, c + 1, 0); } if (img_row_fol[c + 1] > 0){ imgLabels_row_fol[c + 1] = iLabel; sopArray_[rowBegin](r + 1, c + 1, iLabel); } else{ imgLabels_row_fol[c + 1] = 0; sopArray_[rowBegin](r + 1, c + 1, 0); } } } else{ imgLabels_row[c] = 0; imgLabels_row_fol[c] = 0; sopArray_[rowBegin](r, c, 0); sopArray_[rowBegin](r + 1, c, 0); if (c + 1 < imgLabels_.cols) { imgLabels_row[c + 1] = 0; imgLabels_row_fol[c + 1] = 0; sopArray_[rowBegin](r, c + 1, 0); sopArray_[rowBegin](r + 1, c + 1, 0); } } } } }// END case 3 else{ //Case 4: nothing odd for (; r < rowEnd; r += 2){ // Get rows pointer const PixelT * const img_row = img_.ptr<PixelT>(r); const PixelT * const img_row_fol = (PixelT *)(((char *)img_row) + img_.step.p[0]); LabelT * const imgLabels_row = imgLabels_.ptr<LabelT>(r); LabelT * const imgLabels_row_fol = (LabelT *)(((char *)imgLabels_row) + imgLabels_.step.p[0]); // Get rows pointer for (int c = 0; c < imgLabels_.cols; c += 2) { LabelT iLabel = imgLabels_row[c]; if (iLabel > 0) { iLabel = P_[iLabel]; if (img_row[c] > 0){ imgLabels_row[c] = iLabel; sopArray_[rowBegin](r, c, iLabel); } else{ imgLabels_row[c] = 0; sopArray_[rowBegin](r, c, 0); } if (img_row[c + 1] > 0){ imgLabels_row[c + 1] = iLabel; sopArray_[rowBegin](r, c + 1, iLabel); } else{ imgLabels_row[c + 1] = 0; sopArray_[rowBegin](r, c + 1, 0); } if (img_row_fol[c] > 0){ imgLabels_row_fol[c] = iLabel; sopArray_[rowBegin](r + 1, c, iLabel); } else{ imgLabels_row_fol[c] = 0; sopArray_[rowBegin](r + 1, c, 0); } if (img_row_fol[c + 1] > 0){ imgLabels_row_fol[c + 1] = iLabel; sopArray_[rowBegin](r + 1, c + 1, iLabel); } else{ imgLabels_row_fol[c + 1] = 0; sopArray_[rowBegin](r + 1, c + 1, 0); } } else { imgLabels_row[c] = 0; imgLabels_row[c + 1] = 0; imgLabels_row_fol[c] = 0; imgLabels_row_fol[c + 1] = 0; sopArray_[rowBegin](r, c, 0); sopArray_[rowBegin](r, c + 1, 0); sopArray_[rowBegin](r + 1, c, 0); sopArray_[rowBegin](r + 1, c + 1, 0); } } }//END case 4 } } } else{ //the first thread uses sop in order to make less merges sop_.setNextLoc(rowEnd); if (imgLabels_.rows& 1){ if (imgLabels_.cols& 1){ //Case 1: both rows and cols odd for (; r < rowEnd; r += 2){ // Get rows pointer const PixelT * const img_row = img_.ptr<PixelT>(r); const PixelT * const img_row_fol = (PixelT *)(((char *)img_row) + img_.step.p[0]); LabelT * const imgLabels_row = imgLabels_.ptr<LabelT>(r); LabelT * const imgLabels_row_fol = (LabelT *)(((char *)imgLabels_row) + imgLabels_.step.p[0]); // Get rows pointer for (int c = 0; c < imgLabels_.cols; c += 2) { LabelT iLabel = imgLabels_row[c]; if (iLabel > 0) { iLabel = P_[iLabel]; if (img_row[c] > 0){ imgLabels_row[c] = iLabel; sop_(r, c, iLabel); } else{ imgLabels_row[c] = 0; sop_(r, c, 0); } if (c + 1 < imgLabels_.cols) { if (img_row[c + 1] > 0){ imgLabels_row[c + 1] = iLabel; sop_(r, c + 1, iLabel); } else{ imgLabels_row[c + 1] = 0; sop_(r, c + 1, 0); } if (r + 1 < imgLabels_.rows) { if (img_row_fol[c] > 0){ imgLabels_row_fol[c] = iLabel; sop_(r + 1, c, iLabel); } else{ imgLabels_row_fol[c] = 0; sop_(r + 1, c, 0); } if (img_row_fol[c + 1] > 0){ imgLabels_row_fol[c + 1] = iLabel; sop_(r + 1, c + 1, iLabel); } else{ imgLabels_row_fol[c + 1] = 0; sop_(r + 1, c + 1, 0); } } } else if (r + 1 < imgLabels_.rows) { if (img_row_fol[c] > 0){ imgLabels_row_fol[c] = iLabel; sop_(r + 1, c, iLabel); } else{ imgLabels_row_fol[c] = 0; sop_(r + 1, c, 0); } } } else { imgLabels_row[c] = 0; sop_(r, c, 0); if (c + 1 < imgLabels_.cols) { imgLabels_row[c + 1] = 0; sop_(r, c + 1, 0); if (r + 1 < imgLabels_.rows) { imgLabels_row_fol[c] = 0; imgLabels_row_fol[c + 1] = 0; sop_(r + 1, c, 0); sop_(r + 1, c + 1, 0); } } else if (r + 1 < imgLabels_.rows) { imgLabels_row_fol[c] = 0; sop_(r + 1, c, 0); } } } } }//END Case 1 else{ //Case 2: only rows odd for (; r < rowEnd; r += 2){ // Get rows pointer const PixelT * const img_row = img_.ptr<PixelT>(r); const PixelT * const img_row_fol = (PixelT *)(((char *)img_row) + img_.step.p[0]); LabelT * const imgLabels_row = imgLabels_.ptr<LabelT>(r); LabelT * const imgLabels_row_fol = (LabelT *)(((char *)imgLabels_row) + imgLabels_.step.p[0]); // Get rows pointer for (int c = 0; c < imgLabels_.cols; c += 2) { LabelT iLabel = imgLabels_row[c]; if (iLabel > 0) { iLabel = P_[iLabel]; if (img_row[c] > 0){ imgLabels_row[c] = iLabel; sop_(r, c, iLabel); } else{ imgLabels_row[c] = 0; sop_(r, c, 0); } if (img_row[c + 1] > 0){ imgLabels_row[c + 1] = iLabel; sop_(r, c + 1, iLabel); } else{ imgLabels_row[c + 1] = 0; sop_(r, c + 1, 0); } if (r + 1 < imgLabels_.rows) { if (img_row_fol[c] > 0){ imgLabels_row_fol[c] = iLabel; sop_(r + 1, c, iLabel); } else{ imgLabels_row_fol[c] = 0; sop_(r + 1, c, 0); } if (img_row_fol[c + 1] > 0){ imgLabels_row_fol[c + 1] = iLabel; sop_(r + 1, c + 1, iLabel); } else{ imgLabels_row_fol[c + 1] = 0; sop_(r + 1, c + 1, 0); } } } else { imgLabels_row[c] = 0; imgLabels_row[c + 1] = 0; sop_(r, c, 0); sop_(r, c + 1, 0); if (r + 1 < imgLabels_.rows) { imgLabels_row_fol[c] = 0; imgLabels_row_fol[c + 1] = 0; sop_(r + 1, c, 0); sop_(r + 1, c + 1, 0); } } } } }// END Case 2 } else{ if (imgLabels_.cols& 1){ //Case 3: only cols odd for (; r < rowEnd; r += 2){ // Get rows pointer const PixelT * const img_row = img_.ptr<PixelT>(r); const PixelT * const img_row_fol = (PixelT *)(((char *)img_row) + img_.step.p[0]); LabelT * const imgLabels_row = imgLabels_.ptr<LabelT>(r); LabelT * const imgLabels_row_fol = (LabelT *)(((char *)imgLabels_row) + imgLabels_.step.p[0]); // Get rows pointer for (int c = 0; c < imgLabels_.cols; c += 2) { LabelT iLabel = imgLabels_row[c]; if (iLabel > 0) { iLabel = P_[iLabel]; if (img_row[c] > 0){ imgLabels_row[c] = iLabel; sop_(r, c, iLabel); } else{ imgLabels_row[c] = 0; sop_(r, c, 0); } if (img_row_fol[c] > 0){ imgLabels_row_fol[c] = iLabel; sop_(r + 1, c, iLabel); } else{ imgLabels_row_fol[c] = 0; sop_(r + 1, c, 0); } if (c + 1 < imgLabels_.cols) { if (img_row[c + 1] > 0){ imgLabels_row[c + 1] = iLabel; sop_(r, c + 1, iLabel); } else{ imgLabels_row[c + 1] = 0; sop_(r, c + 1, 0); } if (img_row_fol[c + 1] > 0){ imgLabels_row_fol[c + 1] = iLabel; sop_(r + 1, c + 1, iLabel); } else{ imgLabels_row_fol[c + 1] = 0; sop_(r + 1, c + 1, 0); } } } else{ imgLabels_row[c] = 0; imgLabels_row_fol[c] = 0; sop_(r, c, 0); sop_(r + 1, c, 0); if (c + 1 < imgLabels_.cols) { imgLabels_row[c + 1] = 0; imgLabels_row_fol[c + 1] = 0; sop_(r, c + 1, 0); sop_(r + 1, c + 1, 0); } } } } }// END case 3 else{ //Case 4: nothing odd for (; r < rowEnd; r += 2){ // Get rows pointer const PixelT * const img_row = img_.ptr<PixelT>(r); const PixelT * const img_row_fol = (PixelT *)(((char *)img_row) + img_.step.p[0]); LabelT * const imgLabels_row = imgLabels_.ptr<LabelT>(r); LabelT * const imgLabels_row_fol = (LabelT *)(((char *)imgLabels_row) + imgLabels_.step.p[0]); // Get rows pointer for (int c = 0; c < imgLabels_.cols; c += 2) { LabelT iLabel = imgLabels_row[c]; if (iLabel > 0) { iLabel = P_[iLabel]; if (img_row[c] > 0){ imgLabels_row[c] = iLabel; sop_(r, c, iLabel); } else{ imgLabels_row[c] = 0; sop_(r, c, 0); } if (img_row[c + 1] > 0){ imgLabels_row[c + 1] = iLabel; sop_(r, c + 1, iLabel); } else{ imgLabels_row[c + 1] = 0; sop_(r, c + 1, 0); } if (img_row_fol[c] > 0){ imgLabels_row_fol[c] = iLabel; sop_(r + 1, c, iLabel); } else{ imgLabels_row_fol[c] = 0; sop_(r + 1, c, 0); } if (img_row_fol[c + 1] > 0){ imgLabels_row_fol[c + 1] = iLabel; sop_(r + 1, c + 1, iLabel); } else{ imgLabels_row_fol[c + 1] = 0; sop_(r + 1, c + 1, 0); } } else { imgLabels_row[c] = 0; imgLabels_row[c + 1] = 0; imgLabels_row_fol[c] = 0; imgLabels_row_fol[c + 1] = 0; sop_(r, c, 0); sop_(r, c + 1, 0); sop_(r + 1, c, 0); sop_(r + 1, c + 1, 0); } } }//END case 4 } } } } }; inline static void mergeLabels(const cv::Mat& img, cv::Mat& imgLabels, LabelT *P, int *chunksSizeAndLabels){ // Merge Mask // +---+---+---+ // |P -|Q -|R -| // |- -|- -|- -| // +---+---+---+ // |X -| // |- -| // +---+ const int w = imgLabels.cols, h = imgLabels.rows; for (int r = chunksSizeAndLabels[0]; r < h; r = chunksSizeAndLabels[r]){ LabelT * const imgLabels_row = imgLabels.ptr<LabelT>(r); LabelT * const imgLabels_row_prev_prev = (LabelT *)(((char *)imgLabels_row) - imgLabels.step.p[0] - imgLabels.step.p[0]); const PixelT * const img_row = img.ptr<PixelT>(r); const PixelT * const img_row_prev = (PixelT *)(((char *)img_row) - img.step.p[0]); for (int c = 0; c < w; c += 2){ #define condition_x imgLabels_row[c] > 0 #define condition_pppr c > 1 && imgLabels_row_prev_prev[c - 2] > 0 #define condition_qppr imgLabels_row_prev_prev[c] > 0 #define condition_qppr1 c < w - 1 #define condition_qppr2 c < w #define condition_rppr c < w - 2 && imgLabels_row_prev_prev[c + 2] > 0 if (condition_x){ if (condition_pppr){ //check in img if (img_row[c] > 0 && img_row_prev[c - 1] > 0) //assign the same label imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c - 2], imgLabels_row[c]); } if (condition_qppr){ if (condition_qppr1){ if ((img_row[c] > 0 && img_row_prev[c] > 0) || (img_row[c + 1] > 0 && img_row_prev[c] > 0) || (img_row[c] > 0 && img_row_prev[c + 1] > 0) || (img_row[c + 1] > 0 && img_row_prev[c + 1] > 0)){ imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c]); } } else /*if (condition_qppr2)*/{ if (img_row[c] > 0 && img_row_prev[c] > 0) imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c]); } } if (condition_rppr){ if (img_row[c + 1] > 0 && img_row_prev[c + 2] > 0) imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c]); } } } } } LabelT operator()(const cv::Mat& img, cv::Mat& imgLabels, int connectivity, StatsOp& sop){ CV_Assert(img.rows == imgLabels.rows); CV_Assert(img.cols == imgLabels.cols); CV_Assert(connectivity == 8); const int h = img.rows; const int w = img.cols; //A quick and dirty upper bound for the maximum number of labels. //Following formula comes from the fact that a 2x2 block in 8-connectivity case //can never have more than 1 new label and 1 label for background. //Worst case image example pattern: //1 0 1 0 1... //0 0 0 0 0... //1 0 1 0 1... //............ const size_t Plength = size_t(((h + 1) / 2) * size_t((w + 1) / 2)) + 1; //Array used to store info and labeled pixel by each thread. //Different threads affect different memory location of chunksSizeAndLabels const int chunksSizeAndLabelsSize = roundUp(h, 2); std::vector<int> chunksSizeAndLabels(chunksSizeAndLabelsSize); //Tree of labels std::vector<LabelT> P(Plength, 0); //First label is for background //P[0] = 0; cv::Range range2(0, divUp(h, 2)); const double nParallelStripes = std::max(1, std::min(h / 2, getNumThreads()*4)); //First scan cv::parallel_for_(range2, FirstScan(img, imgLabels, P.data(), chunksSizeAndLabels.data()), nParallelStripes); //merge labels of different chunks mergeLabels(img, imgLabels, P.data(), chunksSizeAndLabels.data()); LabelT nLabels = 1; for (int i = 0; i < h; i = chunksSizeAndLabels[i]){ CV_DbgAssert(i + 1 < chunksSizeAndLabelsSize); flattenL(P.data(), stripeFirstLabel8Connectivity<LabelT>(i, w), chunksSizeAndLabels[i + 1], nLabels); } //Array for statistics data std::vector<StatsOp> sopArray(h); sop.init(nLabels); //Second scan cv::parallel_for_(range2, SecondScan(img, imgLabels, P.data(), sop, sopArray.data(), nLabels), nParallelStripes); StatsOp::mergeStats(imgLabels, sopArray.data(), sop, nLabels); sop.finish(); return nLabels; } };//End struct LabelingGranaParallel //Implementation of BBDT (Block-Based with Decision Tree) algorithm, as described in "Optimized Block-based Connected //Components Labeling with Decision Trees", IEEE Transactions on Image Processing, Costantino Grana et. al. template<typename LabelT, typename PixelT, typename StatsOp = NoOp > struct LabelingGrana{ LabelT operator()(const cv::Mat& img, cv::Mat& imgLabels, int connectivity, StatsOp& sop){ CV_Assert(img.rows == imgLabels.rows); CV_Assert(img.cols == imgLabels.cols); CV_Assert(connectivity == 8); const int h = img.rows; const int w = img.cols; //A quick and dirty upper bound for the maximum number of labels. //Following formula comes from the fact that a 2x2 block in 8-connectivity case //can never have more than 1 new label and 1 label for background. //Worst case image example pattern: //1 0 1 0 1... //0 0 0 0 0... //1 0 1 0 1... //............ const size_t Plength = size_t(((h + 1) / 2) * size_t((w + 1) / 2)) + 1; std::vector<LabelT> P_(Plength, 0); LabelT *P = P_.data(); //P[0] = 0; LabelT lunique = 1; // First scan for (int r = 0; r < h; r += 2) { // Get rows pointer const PixelT * const img_row = img.ptr<PixelT>(r); const PixelT * const img_row_prev = (PixelT *)(((char *)img_row) - img.step.p[0]); const PixelT * const img_row_prev_prev = (PixelT *)(((char *)img_row_prev) - img.step.p[0]); const PixelT * const img_row_fol = (PixelT *)(((char *)img_row) + img.step.p[0]); LabelT * const imgLabels_row = imgLabels.ptr<LabelT>(r); LabelT * const imgLabels_row_prev_prev = (LabelT *)(((char *)imgLabels_row) - imgLabels.step.p[0] - imgLabels.step.p[0]); for (int c = 0; c < w; c += 2) { // We work with 2x2 blocks // +-+-+-+ // |P|Q|R| // +-+-+-+ // |S|X| // +-+-+ // The pixels are named as follows // +---+---+---+ // |a b|c d|e f| // |g h|i j|k l| // +---+---+---+ // |m n|o p| // |q r|s t| // +---+---+ // Pixels a, f, l, q are not needed, since we need to understand the // the connectivity between these blocks and those pixels only matter // when considering the outer connectivities // A bunch of defines used to check if the pixels are foreground, // without going outside the image limits. #define condition_b c-1>=0 && r-2>=0 && img_row_prev_prev[c-1]>0 #define condition_c r-2>=0 && img_row_prev_prev[c]>0 #define condition_d c+1<w && r-2>=0 && img_row_prev_prev[c+1]>0 #define condition_e c+2<w && r-2>=0 && img_row_prev_prev[c+2]>0 #define condition_g c-2>=0 && r-1>=0 && img_row_prev[c-2]>0 #define condition_h c-1>=0 && r-1>=0 && img_row_prev[c-1]>0 #define condition_i r-1>=0 && img_row_prev[c]>0 #define condition_j c+1<w && r-1>=0 && img_row_prev[c+1]>0 #define condition_k c+2<w && r-1>=0 && img_row_prev[c+2]>0 #define condition_m c-2>=0 && img_row[c-2]>0 #define condition_n c-1>=0 && img_row[c-1]>0 #define condition_o img_row[c]>0 #define condition_p c+1<w && img_row[c+1]>0 #define condition_r c-1>=0 && r+1<h && img_row_fol[c-1]>0 #define condition_s r+1<h && img_row_fol[c]>0 #define condition_t c+1<w && r+1<h && img_row_fol[c+1]>0 // This is a decision tree which allows to choose which action to // perform, checking as few conditions as possible. // Actions: the blocks label are provisionally stored in the top left // pixel of the block in the labels image if (condition_o) { if (condition_n) { if (condition_j) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { if (condition_h) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_g) { if (condition_b) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } else { if (condition_p) { if (condition_k) { if (condition_d) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { if (condition_h) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_g) { if (condition_b) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } else { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } } else { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } } } else { if (condition_r) { if (condition_j) { if (condition_m) { if (condition_h) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } else { if (condition_g) { if (condition_b) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } else { if (condition_i) { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } else { if (condition_h) { if (condition_c) { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } else { //Action_14: Merge labels of block P, Q and S imgLabels_row[c] = set_union(P, set_union(P, imgLabels_row_prev_prev[c - 2], imgLabels_row_prev_prev[c]), imgLabels_row[c - 2]); continue; } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } } else { if (condition_p) { if (condition_k) { if (condition_m) { if (condition_h) { if (condition_d) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } else { if (condition_d) { if (condition_g) { if (condition_b) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } else { if (condition_i) { if (condition_g) { if (condition_b) { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } else { //Action_16: labels of block Q, R and S imgLabels_row[c] = set_union(P, set_union(P, imgLabels_row_prev_prev[c], imgLabels_row_prev_prev[c + 2]), imgLabels_row[c - 2]); continue; } } else { //Action_16: labels of block Q, R and S imgLabels_row[c] = set_union(P, set_union(P, imgLabels_row_prev_prev[c], imgLabels_row_prev_prev[c + 2]), imgLabels_row[c - 2]); continue; } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } } else { if (condition_i) { if (condition_d) { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } else { //Action_16: labels of block Q, R and S imgLabels_row[c] = set_union(P, set_union(P, imgLabels_row_prev_prev[c], imgLabels_row_prev_prev[c + 2]), imgLabels_row[c - 2]); continue; } } else { if (condition_h) { if (condition_d) { if (condition_c) { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } else { //Action_15: Merge labels of block P, R and S imgLabels_row[c] = set_union(P, set_union(P, imgLabels_row_prev_prev[c - 2], imgLabels_row_prev_prev[c + 2]), imgLabels_row[c - 2]); continue; } } else { //Action_15: Merge labels of block P, R and S imgLabels_row[c] = set_union(P, set_union(P, imgLabels_row_prev_prev[c - 2], imgLabels_row_prev_prev[c + 2]), imgLabels_row[c - 2]); continue; } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } } else { if (condition_h) { if (condition_m) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { // ACTION_9 Merge labels of block P and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c - 2], imgLabels_row[c - 2]); continue; } } else { if (condition_i) { if (condition_m) { if (condition_g) { if (condition_b) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } } } } else { if (condition_h) { if (condition_m) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { // ACTION_9 Merge labels of block P and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c - 2], imgLabels_row[c - 2]); continue; } } else { if (condition_i) { if (condition_m) { if (condition_g) { if (condition_b) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } } } } } else { if (condition_j) { if (condition_i) { //Action_4: Assign label of block Q imgLabels_row[c] = imgLabels_row_prev_prev[c]; continue; } else { if (condition_h) { if (condition_c) { //Action_4: Assign label of block Q imgLabels_row[c] = imgLabels_row_prev_prev[c]; continue; } else { //Action_7: Merge labels of block P and Q imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c - 2], imgLabels_row_prev_prev[c]); continue; } } else { //Action_4: Assign label of block Q imgLabels_row[c] = imgLabels_row_prev_prev[c]; continue; } } } else { if (condition_p) { if (condition_k) { if (condition_i) { if (condition_d) { //Action_5: Assign label of block R imgLabels_row[c] = imgLabels_row_prev_prev[c + 2]; continue; } else { // ACTION_10 Merge labels of block Q and R imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row_prev_prev[c + 2]); continue; } } else { if (condition_h) { if (condition_d) { if (condition_c) { //Action_5: Assign label of block R imgLabels_row[c] = imgLabels_row_prev_prev[c + 2]; continue; } else { //Action_8: Merge labels of block P and R imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c - 2], imgLabels_row_prev_prev[c + 2]); continue; } } else { //Action_8: Merge labels of block P and R imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c - 2], imgLabels_row_prev_prev[c + 2]); continue; } } else { //Action_5: Assign label of block R imgLabels_row[c] = imgLabels_row_prev_prev[c + 2]; continue; } } } else { if (condition_i) { //Action_4: Assign label of block Q imgLabels_row[c] = imgLabels_row_prev_prev[c]; continue; } else { if (condition_h) { //Action_3: Assign label of block P imgLabels_row[c] = imgLabels_row_prev_prev[c - 2]; continue; } else { //Action_2: New label (the block has foreground pixels and is not connected to anything else) imgLabels_row[c] = lunique; P[lunique] = lunique; lunique = lunique + 1; continue; } } } } else { if (condition_i) { //Action_4: Assign label of block Q imgLabels_row[c] = imgLabels_row_prev_prev[c]; continue; } else { if (condition_h) { //Action_3: Assign label of block P imgLabels_row[c] = imgLabels_row_prev_prev[c - 2]; continue; } else { //Action_2: New label (the block has foreground pixels and is not connected to anything else) imgLabels_row[c] = lunique; P[lunique] = lunique; lunique = lunique + 1; continue; } } } } } } } else { if (condition_s) { if (condition_p) { if (condition_n) { if (condition_j) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { if (condition_h) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_g) { if (condition_b) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } else { if (condition_k) { if (condition_d) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { if (condition_h) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_g) { if (condition_b) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } else { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } } } else { if (condition_r) { if (condition_j) { if (condition_m) { if (condition_h) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } else { if (condition_g) { if (condition_b) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { if (condition_k) { if (condition_d) { if (condition_m) { if (condition_h) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } else { if (condition_g) { if (condition_b) { if (condition_i) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_c) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } else { if (condition_i) { if (condition_m) { if (condition_h) { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } else { if (condition_g) { if (condition_b) { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } else { //Action_16: labels of block Q, R and S imgLabels_row[c] = set_union(P, set_union(P, imgLabels_row_prev_prev[c], imgLabels_row_prev_prev[c + 2]), imgLabels_row[c - 2]); continue; } } else { //Action_16: labels of block Q, R and S imgLabels_row[c] = set_union(P, set_union(P, imgLabels_row_prev_prev[c], imgLabels_row_prev_prev[c + 2]), imgLabels_row[c - 2]); continue; } } } else { //Action_16: labels of block Q, R and S imgLabels_row[c] = set_union(P, set_union(P, imgLabels_row_prev_prev[c], imgLabels_row_prev_prev[c + 2]), imgLabels_row[c - 2]); continue; } } else { //Action_12: Merge labels of block R and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c + 2], imgLabels_row[c - 2]); continue; } } } else { if (condition_i) { if (condition_m) { if (condition_h) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_g) { if (condition_b) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } } else { //Action_11: Merge labels of block Q and S imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row[c - 2]); continue; } } else { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } } } } else { if (condition_j) { //Action_4: Assign label of block Q imgLabels_row[c] = imgLabels_row_prev_prev[c]; continue; } else { if (condition_k) { if (condition_i) { if (condition_d) { //Action_5: Assign label of block R imgLabels_row[c] = imgLabels_row_prev_prev[c + 2]; continue; } else { // ACTION_10 Merge labels of block Q and R imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row_prev_prev[c + 2]); continue; } } else { //Action_5: Assign label of block R imgLabels_row[c] = imgLabels_row_prev_prev[c + 2]; continue; } } else { if (condition_i) { //Action_4: Assign label of block Q imgLabels_row[c] = imgLabels_row_prev_prev[c]; continue; } else { //Action_2: New label (the block has foreground pixels and is not connected to anything else) imgLabels_row[c] = lunique; P[lunique] = lunique; lunique = lunique + 1; continue; } } } } } } else { if (condition_r) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { if (condition_n) { //Action_6: Assign label of block S imgLabels_row[c] = imgLabels_row[c - 2]; continue; } else { //Action_2: New label (the block has foreground pixels and is not connected to anything else) imgLabels_row[c] = lunique; P[lunique] = lunique; lunique = lunique + 1; continue; } } } } else { if (condition_p) { if (condition_j) { //Action_4: Assign label of block Q imgLabels_row[c] = imgLabels_row_prev_prev[c]; continue; } else { if (condition_k) { if (condition_i) { if (condition_d) { //Action_5: Assign label of block R imgLabels_row[c] = imgLabels_row_prev_prev[c + 2]; continue; } else { // ACTION_10 Merge labels of block Q and R imgLabels_row[c] = set_union(P, imgLabels_row_prev_prev[c], imgLabels_row_prev_prev[c + 2]); continue; } } else { //Action_5: Assign label of block R imgLabels_row[c] = imgLabels_row_prev_prev[c + 2]; continue; } } else { if (condition_i) { //Action_4: Assign label of block Q imgLabels_row[c] = imgLabels_row_prev_prev[c]; continue; } else { //Action_2: New label (the block has foreground pixels and is not connected to anything else) imgLabels_row[c] = lunique; P[lunique] = lunique; lunique = lunique + 1; continue; } } } } else { if (condition_t) { //Action_2: New label (the block has foreground pixels and is not connected to anything else) imgLabels_row[c] = lunique; P[lunique] = lunique; lunique = lunique + 1; continue; } else { // Action_1: No action (the block has no foreground pixels) imgLabels_row[c] = 0; continue; } } } } } } // Second scan + analysis LabelT nLabels = flattenL(P, lunique); sop.init(nLabels); if (imgLabels.rows & 1){ if (imgLabels.cols & 1){ //Case 1: both rows and cols odd for (int r = 0; r < imgLabels.rows; r += 2) { // Get rows pointer const PixelT * const img_row = img.ptr<PixelT>(r); const PixelT * const img_row_fol = (PixelT *)(((char *)img_row) + img.step.p[0]); LabelT * const imgLabels_row = imgLabels.ptr<LabelT>(r); LabelT * const imgLabels_row_fol = (LabelT *)(((char *)imgLabels_row) + imgLabels.step.p[0]); for (int c = 0; c < imgLabels.cols; c += 2) { LabelT iLabel = imgLabels_row[c]; if (iLabel > 0) { iLabel = P[iLabel]; if (img_row[c] > 0){ imgLabels_row[c] = iLabel; sop(r, c, iLabel); } else{ imgLabels_row[c] = 0; sop(r, c, 0); } if (c + 1 < imgLabels.cols) { if (img_row[c + 1] > 0){ imgLabels_row[c + 1] = iLabel; sop(r, c + 1, iLabel); } else{ imgLabels_row[c + 1] = 0; sop(r, c + 1, 0); } if (r + 1 < imgLabels.rows) { if (img_row_fol[c] > 0){ imgLabels_row_fol[c] = iLabel; sop(r + 1, c, iLabel); } else{ imgLabels_row_fol[c] = 0; sop(r + 1, c, 0); } if (img_row_fol[c + 1] > 0){ imgLabels_row_fol[c + 1] = iLabel; sop(r + 1, c + 1, iLabel); } else{ imgLabels_row_fol[c + 1] = 0; sop(r + 1, c + 1, 0); } } } else if (r + 1 < imgLabels.rows) { if (img_row_fol[c] > 0){ imgLabels_row_fol[c] = iLabel; sop(r + 1, c, iLabel); } else{ imgLabels_row_fol[c] = 0; sop(r + 1, c, 0); } } } else { imgLabels_row[c] = 0; sop(r, c, 0); if (c + 1 < imgLabels.cols) { imgLabels_row[c + 1] = 0; sop(r, c + 1, 0); if (r + 1 < imgLabels.rows) { imgLabels_row_fol[c] = 0; imgLabels_row_fol[c + 1] = 0; sop(r + 1, c, 0); sop(r + 1, c + 1, 0); } } else if (r + 1 < imgLabels.rows) { imgLabels_row_fol[c] = 0; sop(r + 1, c, 0); } } } } }//END Case 1 else{ //Case 2: only rows odd for (int r = 0; r < imgLabels.rows; r += 2) { // Get rows pointer const PixelT * const img_row = img.ptr<PixelT>(r); const PixelT * const img_row_fol = (PixelT *)(((char *)img_row) + img.step.p[0]); LabelT * const imgLabels_row = imgLabels.ptr<LabelT>(r); LabelT * const imgLabels_row_fol = (LabelT *)(((char *)imgLabels_row) + imgLabels.step.p[0]); for (int c = 0; c < imgLabels.cols; c += 2) { LabelT iLabel = imgLabels_row[c]; if (iLabel > 0) { iLabel = P[iLabel]; if (img_row[c] > 0){ imgLabels_row[c] = iLabel; sop(r, c, iLabel); } else{ imgLabels_row[c] = 0; sop(r, c, 0); } if (img_row[c + 1] > 0){ imgLabels_row[c + 1] = iLabel; sop(r, c + 1, iLabel); } else{ imgLabels_row[c + 1] = 0; sop(r, c + 1, 0); } if (r + 1 < imgLabels.rows) { if (img_row_fol[c] > 0){ imgLabels_row_fol[c] = iLabel; sop(r + 1, c, iLabel); } else{ imgLabels_row_fol[c] = 0; sop(r + 1, c, 0); } if (img_row_fol[c + 1] > 0){ imgLabels_row_fol[c + 1] = iLabel; sop(r + 1, c + 1, iLabel); } else{ imgLabels_row_fol[c + 1] = 0; sop(r + 1, c + 1, 0); } } } else { imgLabels_row[c] = 0; imgLabels_row[c + 1] = 0; sop(r, c, 0); sop(r, c + 1, 0); if (r + 1 < imgLabels.rows) { imgLabels_row_fol[c] = 0; imgLabels_row_fol[c + 1] = 0; sop(r + 1, c, 0); sop(r + 1, c + 1, 0); } } } } }// END Case 2 } else{ if (imgLabels.cols & 1){ //Case 3: only cols odd for (int r = 0; r < imgLabels.rows; r += 2) { // Get rows pointer const PixelT * const img_row = img.ptr<PixelT>(r); const PixelT * const img_row_fol = (PixelT *)(((char *)img_row) + img.step.p[0]); LabelT * const imgLabels_row = imgLabels.ptr<LabelT>(r); LabelT * const imgLabels_row_fol = (LabelT *)(((char *)imgLabels_row) + imgLabels.step.p[0]); for (int c = 0; c < imgLabels.cols; c += 2) { LabelT iLabel = imgLabels_row[c]; if (iLabel > 0) { iLabel = P[iLabel]; if (img_row[c] > 0){ imgLabels_row[c] = iLabel; sop(r, c, iLabel); } else{ imgLabels_row[c] = 0; sop(r, c, 0); } if (img_row_fol[c] > 0){ imgLabels_row_fol[c] = iLabel; sop(r + 1, c, iLabel); } else{ imgLabels_row_fol[c] = 0; sop(r + 1, c, 0); } if (c + 1 < imgLabels.cols) { if (img_row[c + 1] > 0){ imgLabels_row[c + 1] = iLabel; sop(r, c + 1, iLabel); } else{ imgLabels_row[c + 1] = 0; sop(r, c + 1, 0); } if (img_row_fol[c + 1] > 0){ imgLabels_row_fol[c + 1] = iLabel; sop(r + 1, c + 1, iLabel); } else{ imgLabels_row_fol[c + 1] = 0; sop(r + 1, c + 1, 0); } } } else{ imgLabels_row[c] = 0; imgLabels_row_fol[c] = 0; sop(r, c, 0); sop(r + 1, c, 0); if (c + 1 < imgLabels.cols) { imgLabels_row[c + 1] = 0; imgLabels_row_fol[c + 1] = 0; sop(r, c + 1, 0); sop(r + 1, c + 1, 0); } } } } }// END case 3 else{ //Case 4: nothing odd for (int r = 0; r < imgLabels.rows; r += 2) { // Get rows pointer const PixelT * const img_row = img.ptr<PixelT>(r); const PixelT * const img_row_fol = (PixelT *)(((char *)img_row) + img.step.p[0]); LabelT * const imgLabels_row = imgLabels.ptr<LabelT>(r); LabelT * const imgLabels_row_fol = (LabelT *)(((char *)imgLabels_row) + imgLabels.step.p[0]); for (int c = 0; c < imgLabels.cols; c += 2) { LabelT iLabel = imgLabels_row[c]; if (iLabel > 0) { iLabel = P[iLabel]; if (img_row[c] > 0){ imgLabels_row[c] = iLabel; sop(r, c, iLabel); } else{ imgLabels_row[c] = 0; sop(r, c, 0); } if (img_row[c + 1] > 0){ imgLabels_row[c + 1] = iLabel; sop(r, c + 1, iLabel); } else{ imgLabels_row[c + 1] = 0; sop(r, c + 1, 0); } if (img_row_fol[c] > 0){ imgLabels_row_fol[c] = iLabel; sop(r + 1, c, iLabel); } else{ imgLabels_row_fol[c] = 0; sop(r + 1, c, 0); } if (img_row_fol[c + 1] > 0){ imgLabels_row_fol[c + 1] = iLabel; sop(r + 1, c + 1, iLabel); } else{ imgLabels_row_fol[c + 1] = 0; sop(r + 1, c + 1, 0); } } else { imgLabels_row[c] = 0; imgLabels_row[c + 1] = 0; imgLabels_row_fol[c] = 0; imgLabels_row_fol[c + 1] = 0; sop(r, c, 0); sop(r, c + 1, 0); sop(r + 1, c, 0); sop(r + 1, c + 1, 0); } } } }//END case 4 } sop.finish(); return nLabels; } //End function LabelingGrana operator() };//End struct LabelingGrana }//end namespace connectedcomponents //L's type must have an appropriate depth for the number of pixels in I template<typename StatsOp> static int connectedComponents_sub1(const cv::Mat& I, cv::Mat& L, int connectivity, int ccltype, StatsOp& sop){ CV_Assert(L.channels() == 1 && I.channels() == 1); CV_Assert(connectivity == 8 || connectivity == 4); CV_Assert(ccltype == CCL_SPAGHETTI || ccltype == CCL_BBDT || ccltype == CCL_SAUF || ccltype == CCL_BOLELLI || ccltype == CCL_GRANA || ccltype == CCL_WU || ccltype == CCL_DEFAULT); int lDepth = L.depth(); int iDepth = I.depth(); const char *currentParallelFramework = cv::currentParallelFramework(); const int nThreads = cv::getNumThreads(); CV_Assert(iDepth == CV_8U || iDepth == CV_8S); //Run parallel labeling only if the rows of the image are at least twice the number of available threads const bool is_parallel = currentParallelFramework != NULL && nThreads > 1 && L.rows / nThreads >= 2; if (ccltype == CCL_SAUF || ccltype == CCL_WU || ((ccltype == CCL_BBDT || ccltype == CCL_GRANA) && connectivity == 4)){ // SAUF algorithm is used using connectedcomponents::LabelingWu; using connectedcomponents::LabelingWuParallel; //warn if L's depth is not sufficient? if (lDepth == CV_8U){ //Not supported yet } else if (lDepth == CV_16U){ return (int)LabelingWu<ushort, uchar, StatsOp>()(I, L, connectivity, sop); } else if (lDepth == CV_32S){ //note that signed types don't really make sense here and not being able to use unsigned matters for scientific projects //OpenCV: how should we proceed? .at<T> typechecks in debug mode if (!is_parallel) return (int)LabelingWu<int, uchar, StatsOp>()(I, L, connectivity, sop); else return (int)LabelingWuParallel<int, uchar, StatsOp>()(I, L, connectivity, sop); } } else if ((ccltype == CCL_BBDT || ccltype == CCL_GRANA) && connectivity == 8){ // BBDT algorithm is used using connectedcomponents::LabelingGrana; using connectedcomponents::LabelingGranaParallel; //warn if L's depth is not sufficient? if (lDepth == CV_8U){ //Not supported yet } else if (lDepth == CV_16U){ return (int)LabelingGrana<ushort, uchar, StatsOp>()(I, L, connectivity, sop); } else if (lDepth == CV_32S){ //note that signed types don't really make sense here and not being able to use unsigned matters for scientific projects //OpenCV: how should we proceed? .at<T> typechecks in debug mode if (!is_parallel) return (int)LabelingGrana<int, uchar, StatsOp>()(I, L, connectivity, sop); else return (int)LabelingGranaParallel<int, uchar, StatsOp>()(I, L, connectivity, sop); } } else if (ccltype == CCL_SPAGHETTI || ccltype == CCL_BOLELLI || ccltype == CCL_DEFAULT) { // Spaghetti algorithm is used if (connectivity == 8) { using connectedcomponents::LabelingBolelli; using connectedcomponents::LabelingBolelliParallel; //warn if L's depth is not sufficient? if (lDepth == CV_8U) { //Not supported yet } else if (lDepth == CV_16U) { return (int)LabelingBolelli<ushort, uchar, StatsOp>()(I, L, connectivity, sop); } else if (lDepth == CV_32S) { //note that signed types don't really make sense here and not being able to use unsigned matters for scientific projects //OpenCV: how should we proceed? .at<T> typechecks in debug mode if (!is_parallel) return (int)LabelingBolelli<int, uchar, StatsOp>()(I, L, connectivity, sop); else return (int)LabelingBolelliParallel<int, uchar, StatsOp>()(I, L, connectivity, sop); } } else { using connectedcomponents::LabelingBolelli4C; using connectedcomponents::LabelingBolelli4CParallel; //warn if L's depth is not sufficient? if (lDepth == CV_8U) { //Not supported yet } else if (lDepth == CV_16U) { return (int)LabelingBolelli4C<ushort, uchar, StatsOp>()(I, L, connectivity, sop); } else if (lDepth == CV_32S) { //note that signed types don't really make sense here and not being able to use unsigned matters for scientific projects //OpenCV: how should we proceed? .at<T> typechecks in debug mode if (!is_parallel) return (int)LabelingBolelli4C<int, uchar, StatsOp>()(I, L, connectivity, sop); else return (int)LabelingBolelli4CParallel<int, uchar, StatsOp>()(I, L, connectivity, sop); } } } CV_Error(CV_StsUnsupportedFormat, "unsupported label/image type"); } } // Simple wrapper to ensure binary and source compatibility (ABI) int cv::connectedComponents(InputArray img_, OutputArray _labels, int connectivity, int ltype){ return cv::connectedComponents(img_, _labels, connectivity, ltype, CCL_DEFAULT); } int cv::connectedComponents(InputArray img_, OutputArray _labels, int connectivity, int ltype, int ccltype){ const cv::Mat img = img_.getMat(); _labels.create(img.size(), CV_MAT_DEPTH(ltype)); cv::Mat labels = _labels.getMat(); connectedcomponents::NoOp sop; if (ltype == CV_16U){ return connectedComponents_sub1(img, labels, connectivity, ccltype, sop); } else if (ltype == CV_32S){ return connectedComponents_sub1(img, labels, connectivity, ccltype, sop); } else{ CV_Error(CV_StsUnsupportedFormat, "the type of labels must be 16u or 32s"); } } // Simple wrapper to ensure binary and source compatibility (ABI) int cv::connectedComponentsWithStats(InputArray img_, OutputArray _labels, OutputArray statsv, OutputArray centroids, int connectivity, int ltype) { return cv::connectedComponentsWithStats(img_, _labels, statsv, centroids, connectivity, ltype, CCL_DEFAULT); } int cv::connectedComponentsWithStats(InputArray img_, OutputArray _labels, OutputArray statsv, OutputArray centroids, int connectivity, int ltype, int ccltype) { const cv::Mat img = img_.getMat(); _labels.create(img.size(), CV_MAT_DEPTH(ltype)); cv::Mat labels = _labels.getMat(); connectedcomponents::CCStatsOp sop(statsv, centroids); if (ltype == CV_16U){ return connectedComponents_sub1(img, labels, connectivity, ccltype, sop); } else if (ltype == CV_32S){ return connectedComponents_sub1(img, labels, connectivity, ccltype, sop); } else{ CV_Error(CV_StsUnsupportedFormat, "the type of labels must be 16u or 32s"); return 0; } }
三、解决方案
先看看chatGPT的回答:
- Make sure that you have the latest version of OpenCV installed.
- Try using a different algorithm for connected components. For example, you can use the
cv::connectedComponents
function with 4-connectivity instead ofcv::connectedComponentsWithStats
. - If you need to use
cv::connectedComponentsWithStats
with 4-connectivity, you can try disabling the parallel implementation by setting thecv::CCL_DEFAULT
flag. For example:cv::Mat labels, stats, centroids; cv::connectedComponentsWithStats(image, labels, stats, centroids, 4, CV_32S, cv::CCL_DEFAULT);
- If disabling the parallel implementation does not solve the problem, you can try reducing the size of the input image. The error message may be related to a memory allocation issue, and reducing the size of the image may help to avoid the error.
- Finally, if none of the above steps work, you can try reporting the issue to the OpenCV community. They may be able to provide additional guidance or fix the issue in a future release.
其实,我们已经知道了出错的原因,这里解决方案也很明确,在我们的环境中,阻止并行实现可以解决问题,但是上面chatGPT给出的描述不太准确,对照源代码,我们容易看出,lDepth应该用CV_16U才是正解(亲测有效,但是需要注意结果的差别):