Why do we use the center pixel of the closest patches versus a random pixel in the patch as a candidate for p?
GKohavi
I think if we were to do that, it would mostly be the same, hover the idea of getting pixels that have similar contexts is a little more intuitive and simple to implement than taking a random pixel from a similar patch.
Why do we use the center pixel of the closest patches versus a random pixel in the patch as a candidate for p?
I think if we were to do that, it would mostly be the same, hover the idea of getting pixels that have similar contexts is a little more intuitive and simple to implement than taking a random pixel from a similar patch.