OpenCV Mat::data is zero after image operation

Question

I have a problem in OpenCV C++. I read in an image from file and use the threshold-function to get a binary image. After that I want to access the pixel values of the thresholded image.

PROBLEM: The Mat::data array of the thresholded image is 0.

I use threshold(src, img, 220, 255, CV_THRESH_BINARY); with src and img beig cv::Mat.

After that call in the img::data = " "

Here is my code:

// read image source    
Mat src = imread("image.png", CV_LOAD_IMAGE_COLOR);
if(src.empty())
{   
    return -1;
}

cvtColor(src,src,CV_RGB2GRAY);

threshold(src, src, 220, 255, CV_THRESH_BINARY );

    int line[1200];
int lineCount = 0;

for(int i=src.rows; i>src.rows/2; i-=10)    
{
    uchar* rowi = src.ptr(i-1);

    int columnCount = 0;
    // begin on the left of the image
    for(int j=0; j<src.cols; j++)
    {
        uchar grayscale = src.at<uchar>(rowi[j]);   

        // write data to array
        line[columnCount] = grayscale;

        columnCount++;
    }
}

I think the problem might be about cv::Mat with referencing and copying Mat::data:

Thanks for your help!

Answer 1

I suppose you have checked that your image.png (after having been converted to grayscale) contains values higher than 220?

You may want to try to perform cvtColor as a non-in-place operation, as it might not properly support in-place conversion (see https://stackoverflow.com/a/15345586/19254 ). This you would do for example like this:

Mat dst;
cvtColor(src, dst, CV_RGB2GRAY);

... and then do threshold on dst . Alternatively, you could read the image into memory in grayscale mode to begin with (second parameter to imread can be CV_LOAD_IMAGE_GRAYSCALE ), in which case the cvtColor call is not necessary.

Between each call, you can imwrite your image to disk in order to find out which step is not working as expected.