求助复杂背景下区域提取方法

[victor60a]

如图，背景比较复杂，周围的划痕为正常区域，希望可以提取出中间黑色的小块区域，试过频域滤波，和几种阈值分割方法，效果都比较一般，提取出来的区域会带有周围的划痕，求助有没有方法能只提取中间的黑色小块区域，谢啦

[赵4老师]

#include "opencv2/highgui/highgui.hpp"
#include "opencv2/imgproc/imgproc.hpp"
#include "opencv2/imgproc/imgproc_c.h"
using namespace std;
using namespace cv;
Mat img,smallImg,gray,bw;
vector<Vec4i> hierarchy;
vector<vector<Point> > contours;
int threshval=128;
Rect r;
Rect maxrect,brect;
int idx,n;
const static Scalar colors[15]={
    CV_RGB(  0,  0,128),
    CV_RGB(  0,128,  0),
    CV_RGB(  0,128,128),
    CV_RGB(128,  0,  0),
    CV_RGB(128,  0,128),
    CV_RGB(128,128,  0),
    CV_RGB(128,128,128),
    CV_RGB(160,160,160),
    CV_RGB(  0,  0,255),
    CV_RGB(  0,255,  0),
    CV_RGB(  0,255,255),
    CV_RGB(255,  0,  0),
    CV_RGB(255,  0,255),
    CV_RGB(255,255,  0),
    CV_RGB(255,255,255),
};
Scalar color;
void gamma_correct(Mat& img, Mat& dst, double gamma) {
	Mat	temp;
	CvMat tmp;

	img.convertTo(temp,	CV_32FC1, 1.0/255.0, 0.0);
	tmp=temp;
	cvPow(&tmp,	&tmp, gamma);
	temp.convertTo(dst , CV_8UC1 , 255.0	, 0.0);
}
int main() {
    cvNamedWindow("display",1);
    img=imread("image.jpg",1);
    r.x=img.cols/10;
    r.y=img.rows/3;
    r.width=img.cols*8/10;
    r.height=img.rows*2/3;
    smallImg=img(r);
    cvtColor(smallImg,gray,CV_BGR2GRAY);
//  medianBlur(gray,gray,5);
    equalizeHist(gray,gray);
    gamma_correct(gray,gray,4.0);
    imshow("display",gray);
    waitKey(0);

	bw=(gray>threshval);
    imshow("display",bw);
    waitKey(0);

	Mat	Structure0=getStructuringElement(MORPH_RECT,Size(3,3));
    erode(bw,bw,Structure0,Point(-1,-1));
	Mat	Structure1=getStructuringElement(MORPH_RECT,Size(6,6));
    dilate(bw,bw,Structure1, Point(-1,-1));
    imshow("display",bw);
    waitKey(0);

    findContours(bw,contours,hierarchy,RETR_EXTERNAL,CHAIN_APPROX_SIMPLE);
    if (!contours.empty()&&!hierarchy.empty()) {
        idx=0;
        n=0;
        vector<Point> approx;
        for (;idx>=0;idx=hierarchy[idx][0]) {
            color=colors[idx%15];
//          drawContours(smallImg,contours,idx,color,1,8,hierarchy);
            approxPolyDP(Mat(contours[idx]), approx, arcLength(Mat(contours[idx]), true)*0.005, true);//0.005为将毛边拉直的系数
            const Point* p = &approx[0];
            int m=(int)approx.size();
            polylines(smallImg, &p, &m, 1, true, color);
            circle(smallImg,Point(p[0].x,p[0].y),3,color);
            circle(smallImg,Point(p[1].x,p[1].y),2,color);
			for	(int i=2;i<m;i++) circle(smallImg,Point(p[i].x,p[i].y),1,color);
            n++;
            if (1==n) {
                maxrect=boundingRect(Mat(contours[idx]));
            } else {
                brect=boundingRect(Mat(contours[idx]));
                CvRect mr(maxrect),br(brect);
                maxrect=cvMaxRect(&mr,&br);
            }
        }
        circle(smallImg,Point(maxrect.x+maxrect.width/2,maxrect.y+maxrect.height/2),2,CV_RGB(255,0,0));
    }
    imshow("display",smallImg);
    waitKey(0);
    cvDestroyWindow("display");
    return 0;
}

[chen_JADE]

使用opencv

[power_mcu]

应该还是比较好处理的。把圆形区域先抠出来，只处理那块就简单了。