在opencv中获取灰度图像中的最大像素值

Question

After I do some image manipulation, and apply mask, I get what I want. I can clearly see on imshow result of "crop" that there's gray pixels in the middle of image. I'm trying to get the maximum pixel value location. I've checked the crop.channels(), which returns 1.

    Mat mask = drawing2;
    drawContours(mask, contours, -1, Scalar(255), CV_FILLED);
    Mat dist;
    distanceTransform( cannyInv, dist, CV_DIST_L2, 3 );
    normalize(dist,dist,0.0,1.0,NORM_MINMAX);
    Mat crop;
    dist.copyTo(crop, mask);
    cout << "max.. "<< *std::max_element(crop.begin<double>(),crop.end<double>()) <<endl;

which returns max.. 4.25593e-08

    for(int y = 0; y < crop.rows; y++)
    {
        for(int x = 0; x < crop.cols; x++)
        {
            if (crop.at<unsigned char>(x,y) > 0){      
                cout << "X........"<<x<<" Y......"<<y<< " = "<<crop.at<unsigned char>(x,y) <<endl;
            }
        }
    }

The output is:

X........604 Y......479 = ¿
X........607 Y......479 =   
X........610 Y......479 = ¿

Help me please

PD: I know that there's similar question. But this is specific problem.

Answer 1

I'm not sure how I solved it. A lot of time has passed. But the code that currently I have and it works is this:

    Mat dist=Mat::zeros(480,640, CV_8UC1);;
    distanceTransform( cannyInv, dist, CV_DIST_L2, 3 );
    Mat distNorm;
    dist.convertTo(distNorm, CV_8UC1,1,0);
    Mat result= Mat::zeros(480,640, CV_8UC1);
    distNorm.copyTo(result, mask);
    Mat tmp=Mat::zeros(480,640, CV_8UC1);
    Mat fik=Mat::zeros(480,640, CV_8UC3);
    for(int i = 0; i < result.rows; i++)
    {
        for(int j = 0; j < result.cols; j++)
        {
            if ( result.at< uchar >( i,j ) > 0){
                uchar val = result.at< uchar >( i,j  );
                if(val>maxVal){
                    if(val>0){
                        cv::circle(tmp,cvPoint(j,i),2,255,-1);
                    }
                    maxVal=val;
                    maxX = j;
                    maxY = i;
                }
            }
        }
    }

Answer 2

Are you sure that normalizing the Mat automatically converts it from uchar to double? It's very likely the data is still stored as uchars and you're reading wrong numbers from it.

Try dist.convertTo(dist, CV_64F); Print the numbers as doubles everywhere

OR work only with uchars.

Answer 3

Try this code:

cout << "X........"
     << x
     << " Y......"
     << y
     << " = "
     << (double) crop.at< unsigned char>(x,y) <<endl;