基于OpenCV的形态学操作的实现
我不做图像处理,却因为选了计算机视觉课而学了一点点的数学形态学。个人感觉形态学应该很有用处,我也看到它在图像处理中的应用之广泛。于是乎基于OpenCV平台做了一点小实验,看了看形态学操作的效果。
一、膨胀、腐蚀、开和闭、形态梯度、顶帽变换(注意代码中i和L的区别,下面是腐蚀的代码,其它类似)
IplImage*oldImg=
cvLoadImage(filePath,CV_LOAD_IMAGE_GRAYSCALE);//导入图像
IplImage* fushiImg =
cvCreateImage(cvGetSize(oldImg),oldImg->depth,1);
IplConvKernel*element=
cvCreateStructuringElementEx( m_struCols,m_struRows, m_struCols/2,m_struRows/2,m_nStruStyle);//创建结构元素
cvErode( oldImg, fushiImg,element,m_nCishu);//腐蚀图像
cvNamedWindow("腐蚀图");
cvShowImage("腐蚀图", fushiImg);
cvWaitKey(0);
cvReleaseImage(&oldImg);
cvReleaseImage(&fushiImg);
cvDestroyWindow("腐蚀图");
二、颗粒分析(计算颗粒数量、求出平均颗粒面积及最大颗粒面积)
IplImage *src = 0; //定义源图像指针
IplImage *tmp = 0; //定义临时图像指针
IplImage *src_back = 0; //定义源图像背景指针
IplImage *dst_gray = 0; //定义源文件去掉背景后的目标灰度图像指针
IplImage *dst_bw = 0; //定义源文件去掉背景后的目标二值图像指针
IplImage *dst_contours = 0; //定义轮廓图像指针
IplConvKernel *element = 0; //定义形态学结构指针
int Number_Object =0; //定义目标对象数量
int contour_area_tmp = 0; //定义目标对象面积临时寄存器
int contour_area_sum = 0; //定义目标所有对象面积的和
int contour_area_ave = 0; //定义目标对象面积平均值
int contour_area_max = 0; //定义目标对象面积最大值
CvMemStorage *stor = 0;
CvSeq * cont = 0;
CvContourScanner contour_scanner;
CvSeq * a_contour= 0;
//1.读取和显示图像
/* the first command line parameter must be image file name */
if ( argc == 2 && (src = cvLoadImage(argv[1], -1))!=0 )
{
;
}
else
{
src = cvLoadImage("rice.jpg", 0);
}
cvNamedWindow( "src", CV_WINDOW_AUTOSIZE );
cvShowImage( "src", src );
//cvSmooth(src, src, CV_MEDIAN, 3, 0, 0, 0); //中值滤波,消除小的噪声;
//2.估计图像背景
tmp = cvCreateImage( cvGetSize(src), src->depth, src->nChannels);
src_back = cvCreateImage( cvGetSize(src), src->depth, src->nChannels);
//创建结构元素
element = cvCreateStructuringElementEx( 4, 4, 1, 1, CV_SHAPE_ELLIPSE, 0);
//用该结构对源图象进行数学形态学的开操作后,估计背景亮度
cvErode( src, tmp, element, 10);
cvDilate( tmp, src_back, element, 10);
cvNamedWindow( "src_back", CV_WINDOW_AUTOSIZE );
cvShowImage( "src_back", src_back );
//3.从源图象中减区背景图像
dst_gray = cvCreateImage( cvGetSize(src), src->depth, src->nChannels);
cvSub( src, src_back, dst_gray, 0);
cvNamedWindow( "dst_gray", CV_WINDOW_AUTOSIZE );
cvShowImage( "dst_gray", dst_gray );
//4.使用阀值操作将图像转换为二值图像
dst_bw = cvCreateImage( cvGetSize(src), src->depth, src->nChannels);
cvThreshold( dst_gray, dst_bw ,50, 255, CV_THRESH_BINARY ); //取阀值为50把图像转为二值图像
//cvAdaptiveThreshold( dst_gray, dst_bw, 255, CV_ADAPTIVE_THRESH_MEAN_C, CV_THRESH_BINARY, 3, 5 );
cvNamedWindow( "dst_bw", CV_WINDOW_AUTOSIZE );
cvShowImage( "dst_bw", dst_bw );
//5.检查图像中的目标对象数量
stor = cvCreateMemStorage(0);
cont = cvCreateSeq(CV_SEQ_ELTYPE_POINT, sizeof(CvSeq), sizeof(CvPoint), stor);
Number_Object = cvFindContours( dst_bw, stor, &cont, sizeof(CvContour),
CV_RETR_TREE, CV_CHAIN_APPROX_SIMPLE, cvPoint(0,0) ); //找到所有轮廓
printf("Number_Object: %d\n", Number_Object);
//6.计算图像中对象的统计属性
dst_contours = cvCreateImage( cvGetSize(src), src->depth, src->nChannels);
cvThreshold( dst_contours, dst_contours ,0, 255, CV_THRESH_BINARY ); //在画轮廓前先把图像变成白色
for(;cont;cont = cont->h_next)
{
cvDrawContours( dst_contours, cont, CV_RGB(255, 0, 0), CV_RGB(255, 0, 0), 0, 1, 8, cvPoint(0, 0) ); //绘制当前轮廓
contour_area_tmp = fabs(cvContourArea( cont, CV_WHOLE_SEQ )); //获取当前轮廓面积
if( contour_area_tmp > contour_area_max )
{
contour_area_max = contour_area_tmp; //找到面积最大的轮廓
}
contour_area_sum += contour_area_tmp; //求所有轮廓的面积和
}
contour_area_ave = contour_area_sum/ Number_Object; //求出所有轮廓的平均值
printf("contour_area_ave: %d\n", contour_area_ave );
printf("contour_area_max: %d\n", contour_area_max );
cvNamedWindow( "dst_contours", CV_WINDOW_AUTOSIZE );
cvShowImage( "dst_contours", dst_contours );
cvWaitKey(-1); //等待退出
cvReleaseImage(&src);
cvReleaseImage(&tmp);
cvReleaseImage(&src_back);
cvReleaseImage(&dst_gray);
cvReleaseImage(&dst_bw);
cvReleaseImage(&dst_contours);
cvReleaseMemStorage(&stor);
cvDestroyWindow( "src" );
cvDestroyWindow( "src_back" );
cvDestroyWindow( "dst_gray" );
cvDestroyWindow( "dst_bw" );
cvDestroyWindow( "dst_contours" );
//void cvDestroyAllWindows(void);
至于实验结果就不贴上来了,呵,都是最最基本的,不贴也罢!
posted on 2009-04-20 22:17 Martin_xiao 阅读(874) 评论(0) 编辑 收藏 举报