Do you achieve anything different if you blur each channel independently?

asked 2018-09-07 01:17:46 -0500

sazr gravatar image

I saw in a tutorial (cant remember where the tutorial was) where someone split a 3 channel image (BGR) into its single channels, median blurred each single channel, performed Canny edge detection on each blurred channel image then joined the 3 resulting edges together. I was thinking, why not just median blur the 3 channel image then use Canny to get the edges instead of splitting first?

My question is; is there any significant difference in results by doing things their way vs mine? Does it do something different that I'm not aware of? Maybe it does something useful I need to learn.

*I'm pretty sure the tutorial was a PyImageSearch one but I cannot find which one.

edit retag flag offensive close merge delete



canny needs one channel image. Splitting image at first step or second step IMHO there is no difference

LBerger gravatar imageLBerger ( 2018-09-07 01:34:15 -0500 )edit

^^ ah, right. Sobel and Laplacian DO work on multichannel images, but not Canny.

berak gravatar imageberak ( 2018-09-07 04:00:22 -0500 )edit