3

As I know, if we consider a 3*3 kernel, we should add a padding of 1px to the source image(if we want to have effect on whole of the image), then we start to put the kernel in upper-left side of the image and multiplying each element of kernel to corresponding pixel on image. Then we sum all the results and put it on the anchor point of kernel(usually center element). Then we should shift the kernel one step to right side and do these things again.

If I am right till here, I have a question about the summation results. I want to know: should we consider the replaced value of image in previously calculated summation and replaced in anchor point in new step of calculation or not?

I mean we must put the anchor point's result in source image and consider it in calculations of shifted kernel? Or we must put it in distance image and we don't consider these results when we shift the kernel on source image(It means don't replace the results on source image for next steps calculations)?

user3486308
  • 285
  • 1
  • 5

1 Answers1

3

Assuming that you are performing a normal discrete convolution, then you would use the original source image for all calculations, and only replace at the end. As you put it, like this:

put it in distance image and we don't consider these results when we shift the kernel on source image(It means don't replace the results on source image for next steps calculations)?

The other option, changing in-place during the operation, does not relate to any common use that I know of, but might have some interesting behaviour in studies of cellular automata or similar.

Neil Slater
  • 32,068
  • 3
  • 43
  • 64