Question

I am trying to access openCV matrix element which comes from converting video frame into grayscale. My code is like this:

int C = 3; 
cv::Mat grayPrevious;
cv::cvtColor ( previousFrame, grayPrevious, CV_RGB2GRAY  );
double* grayPtr = grayPrevious.ptr <double> ();

double mean [600][800][3];

///... // set some values to mean here

for ( int i = 0; i < 600; i ++ )
{
    for ( int j = 0; j < 800; j ++ )
    {
        for ( int m = 0; m < C; m ++ )
        {
            diff [i][j][m]= abs(grayPtr[ 800* (i-1)+j] - mean[i][j][m]);
        }

    }
}

But it crashes in the for loop. Is this something related to datatypes? How can I change it?

Besides, my second question is, I tried to see the elements of grayPrevious matrix on the screen using “grayPrevious.at”, but it shows values like ” -5.235125e+30″ for “double”, or “-11242433” for “int”. This is also confusing me, I was expecting to see values between 0 and 255.

Answer 1

You should use uchar instead of int or double since you are dealing with grayscale images. (You can cast the uchar to print: cout<<(int)grayPrevious.at<uchar>(i,j);)

Is there any reason you are trying to access pixels directly “the hard way”?
You could do grayprevious.at<uchar>(i,j) (assuming your image is 800×600) which is easier to debug and then use direct pixel access.

Edit: I think it should be:

uchar* grayPtr = grayPrevious.ptr <uchar> (); and then grayPtr[grayPrevious.cols*i+j];

You could also do grayPrevious.data[grayPrevious.cols*i+j]; (Still direct access without ptr)