Vavo's profile - activity

It is documented that if I want to use EMD with only weights stored as columns in the signatures I can do that given that I also give a user defined cost matrix. Well, my problem is that it doesn't really work, the function crashes if I try to do that. Are we sure that is is indeed supported?

I am writing a function that uses BLUR with variable padding, however the function crashes if I use BORDER_WRAP (works fine for other padding types). Is this by design or a code issue? Has anyone encountered this problem or should I check my code?

Yes, the link goes to the standard algorithm of NL-Means denoising. There is nothing that accounts for the "fast" keyword or the optimizations that openCV has.

The documentation says that opencv uses the NL-Means denoising algorithm with several computational optimizations. Can anyone tell me what those optimizations are?