So I might be missing something, but it seems the only way to iterate over a cv::UMat is first convert it to a cv::Mat. Is this true? I need to write many functions to operate on individual pixels. I've templatized a lot my code, but things are getting a bit convoluted because of all the switching I have to do. If I didn't have to call mat.getMat(flag) all over the place, that would be much better.
Thanks!
For example, I've implemented a cumulative sum:
cv::Mat CumSum(const cv::Mat& histogram) {
cv::Mat cumsum =
cv::Mat::zeros(histogram.rows, histogram.cols, histogram.type());
cumsum.at<float>(0) = (histogram.at<float>(0));
for (int i = 1; i < histogram.total(); ++i) {
cumsum.at<float>(i) = cumsum.at<float>(i - 1) + histogram.at<float>(i);
}
return cumsum;
}
the at<type> doesn't exist for cv::UMat so I was forced to used function overloading instead of templatizing the in put types.
