Hello everybody!
I am currently using OpenCV-3.1 and I was wondering whether there was a way of indexing a `cv::cuda::GpuMat` channels without calling `cv::cuda::split`.
I don't need to write to these channels, I just need to perform arithmetic operations with each channel separately and save them somewhere else (i.e., `cv::cuda::divide(gpu_red, gpu_blue, gpu_rdb`) ).
I don't want to use split because it copies matrices and it's very computationally expensive (it is currently, by far, my app's bottleneck).
Any ideas would be appreciated.
Thanks!
↧