Revision history [back]

For each image type you have the according data type: uchar for CV_8U, Vec3b for CV_8UC3, ushort for CV_16U, int for CV_32S, float for CV_32F etc.

Always use the according data type:

Mat image(100,100,CV_8UC1);
uchar pixel=image.at<uchar>(10,10);

(you don't need a second type cast: pixel=(uchar)img.at<uchar>(10,10);)

The pixel access in OpenCV is quite well explained in the docs.

It's also good to learn the pointer (line) access.