图像处理基本算法 链码 边界跟踪
链码在图像提取的后期即模式识别是一个很重要的特征,比如进行数字识别或者文字识别都会用到链码的特征,而链码的提取则可以借助于边界跟踪算法获取边界序列,注意是边界序列而不是边界,边界很容易获取,但是要想把边界的点按照一定的顺序输出则要费些功夫。下面采用边界跟踪算法获取边界,并存储在堆栈中,(这里的堆栈实际是C++容器类,是虚拟堆栈)。
利用点的八邻域信息,选择下一个点作为边界点,这个算法需要选择一个开始点,可以选择图像上是目标点,在最上,最左的点。然后查看它的八邻域的点,从右下方45°的位置开始寻找,如果是目标点,将沿顺时针90°作为下一次寻找的方向,如果不是,则逆时针45°继续寻找,一旦找到重复上面的过程。
![](//img-my.csdn.net/uploads/201202/11/0_1328947275Tat9.gif)
利用点的八邻域信息,选择下一个点作为边界点,这个算法需要选择一个开始点,可以选择图像上是目标点,在最上,最左的点。然后查看它的八邻域的点,从右下方45°的位置开始寻找,如果是目标点,将沿顺时针90°作为下一次寻找的方向,如果不是,则逆时针45°继续寻找,一旦找到重复上面的过程。
具体的步骤在算法中有讲解。
/************************************************************************/ /* 查找物体的边界,输出已排序的边界序列 适应于单一区域 */ /************************************************************************/ //若能够输出边界点的序列则是比较有用的 #include<cv.h> #include <highgui.h> #include <iostream> #include <stack> using namespace std; int main(){ IplImage * image,*image2; image = cvLoadImage("E:\\image\\mapleleaf.tif",0); cvNamedWindow("image",1); cvShowImage("image",image); image2 = cvCreateImage(cvSize(image->width, image->height),image->depth,1); cvZero(image2);//image2 赋值为0 //寻找区域的左上角点 CvPoint startPoint = cvPoint(0,0); bool bFindStartpoint = false; int i ,j; unsigned char * ptr,*dst; stack<int> board;//奇数位存储x坐标,偶数位存储y坐标 //当前扫描点 CvPoint currentPoint = cvPoint(0,0); //邻域的8个点的方向 int directions[8][2] = {{0,1},{1,1},{1,0},{1,-1},{0,-1},{-1,-1},{-1,0},{-1,1}}; int beginDirection = 0; bool bFindBoardpoint = false;//寻找到邻域的边界点的判定 for (i = 0 ; i< image->height && bFindStartpoint == false; i++) { for (j = 0 ; j< image->width && bFindStartpoint == false; j++) { ptr = (unsigned char *)(image->imageData + i*image->widthStep + j); if (*ptr == 255) { startPoint.x = j; startPoint.y = i; bFindStartpoint = true; //cout<<"x: " << j <<"y : " <<i <<endl; } } } //进行边界跟踪 每次搜索8个方向的点 找到了即停止 currentPoint = startPoint; bFindStartpoint = false; beginDirection = 0; board.push(startPoint.x); board.push(startPoint.y); while (!bFindStartpoint) { bFindBoardpoint = false; //在8个方向寻找符合条件的边界点 while (!bFindBoardpoint) { //进行出界判定 不对啊 这张图不可能出界啊 ptr = (unsigned char *)(image->imageData + (currentPoint.y + directions[beginDirection][1])* image->widthStep + currentPoint.x + directions[beginDirection][0]); if (*ptr == 255) { bFindBoardpoint = true; currentPoint.x += directions[beginDirection][0]; currentPoint.y += directions[beginDirection][1]; /************************************************************************/ /* 此处添加序列存储的代码 */ /************************************************************************/ //一、将边界存储到图片中 dst = (unsigned char *)image2->imageData + currentPoint.y * image2->widthStep + currentPoint.x; *dst = 255; //二、将边界点的序列存储到一个堆栈中 board.push(currentPoint.x); board.push(currentPoint.y); if (currentPoint.x == startPoint.x && currentPoint.y == startPoint.y ) { bFindStartpoint = true; } //改变下次首先开始扫描的方向 beginDirection -= 2; if (beginDirection < 0) { beginDirection += 8; } } else { beginDirection ++; beginDirection = beginDirection%8; } } //cout<<"currentPoint "<<currentPoint.x <<" "<< currentPoint.y<<endl; } cvNamedWindow("image2",1); cvShowImage("image2",image2); //显示堆栈中的数据 顺时针存储,逆时针显示 //注意:显示时候堆栈中已经没有数据了 /* int x,y; while(!board.empty()) { y = board.top(); board.pop(); x = board.top(); board.pop(); cout<<"x "<<x<<" y "<<y<<endl; } */ cvWaitKey(0); return 0; }
![](http://img-my.csdn.net/uploads/201202/11/0_1328947268vbFT.gif)
![](http://img-my.csdn.net/uploads/201202/11/0_1328947275Tat9.gif)
/************************************************************************/ /* 轮廓跟踪算法获取物体的轮廓序列 生成边界链码 */ /************************************************************************/ #include<cv.h> #include <highgui.h> #include <iostream> #include <stack> using namespace std; int main(){ IplImage * image,*image2,*image3; image = cvLoadImage("E:\\image\\bottle2.tif",0); cvNamedWindow("image",1); cvShowImage("image",image); image2 = cvCreateImage(cvSize(image->width, image->height),image->depth,1); image3 = cvCreateImage(cvSize(image->width, image->height),image->depth,1); cvZero(image2);//image2 赋值为0 cvZero(image3); //寻找区域的左上角点 CvPoint startPoint = cvPoint(0,0); bool bFindStartpoint = false; int i ,j; unsigned char * ptr,*dst; stack<int> board;//奇数位存储x坐标,偶数位存储y坐标 //当前扫描点 CvPoint currentPoint = cvPoint(0,0); //邻域的8个点的方向 int directions[8][2] = {{0,1},{1,1},{1,0},{1,-1},{0,-1},{-1,-1},{-1,0},{-1,1}}; int beginDirection = 0; bool bFindBoardpoint = false;//寻找到邻域的边界点的判定 for (i = 0 ; i< image->height && bFindStartpoint == false; i++) { for (j = 0 ; j< image->width && bFindStartpoint == false; j++) { ptr = (unsigned char *)(image->imageData + i*image->widthStep + j); if (*ptr == 255) { startPoint.x = j; startPoint.y = i; bFindStartpoint = true; //cout<<"x: " << j <<"y : " <<i <<endl; } } } //进行边界跟踪 每次搜索8个方向的点 找到了即停止 currentPoint = startPoint; bFindStartpoint = false; beginDirection = 0; board.push(startPoint.x); board.push(startPoint.y); while (!bFindStartpoint) { bFindBoardpoint = false; //在8个方向寻找符合条件的边界点 while (!bFindBoardpoint) { //进行出界判定 不对啊 这张图不可能出界啊 ptr = (unsigned char *)(image->imageData + (currentPoint.y + directions[beginDirection][1])* image->widthStep + currentPoint.x + directions[beginDirection][0]); if (*ptr == 255) { bFindBoardpoint = true; currentPoint.x += directions[beginDirection][0]; currentPoint.y += directions[beginDirection][1]; /************************************************************************/ /* 此处添加序列存储的代码 */ /************************************************************************/ //一、将边界存储到图片中 dst = (unsigned char *)image2->imageData + currentPoint.y * image2->widthStep + currentPoint.x; *dst = 255; //二、将边界点的序列存储到一个堆栈中 board.push(currentPoint.x); board.push(currentPoint.y); if (currentPoint.x == startPoint.x && currentPoint.y == startPoint.y ) { bFindStartpoint = true; } //改变下次首先开始扫描的方向 beginDirection -= 2; if (beginDirection < 0) { beginDirection += 8; } } else { beginDirection ++; beginDirection = beginDirection%8; } } //cout<<"currentPoint "<<currentPoint.x <<" "<< currentPoint.y<<endl; } cvNamedWindow("image2",1); cvShowImage("image2",image2); //显示堆栈中的数据 顺时针存储,逆时针显示 //注意:显示时候堆栈中已经没有数据了 /* int x,y; while(!board.empty()) { y = board.top(); board.pop(); x = board.top(); board.pop(); cout<<"x "<<x<<" y "<<y<<endl; } */ //Board中存储着边界的序列 转化为8邻域链码,每隔10个点取样 显示 int lianmaLength = (board.size()+5)/10; int* lianma = new int[lianmaLength]; for (i = 0 ; i< lianmaLength && !board.empty();i += 2) { lianma[i+1] = board.top(); board.pop(); lianma[i] = board.top(); board.pop(); for (j = 0; j< 18 && !board.empty();j++) { board.pop(); } } //将数据在image3中显示 int t; for ( t = 0; t < lianmaLength;t += 2) { i = lianma[t+1]; j = lianma[t]; ptr = (unsigned char *)image3->imageData + i*image->widthStep + j; *ptr = 255; } cvNamedWindow("image3",1); cvSaveImage("E:\\image\\bottle2lianma.bmp",image3); cvShowImage("image3",image3); cvWaitKey(0); return 0; }
Keep it simple!