cszhilu1998/SelfDZSR

Question about the chop code

yyknight opened this issue · 1 comments

Thanks for your impressive code. During checking your code, I notice the chop code is little simple. I'm curious how did you remove the borderline during stiching? Especially, it looks like the image is divided into 36 parts.

First, the divided patches are overlapped. Second, the encoder inputs LR (s_c) and Ref (t_c) in the restoration model (see Fig. B in suppl.) are not divided in the testing phase. This operation provides some global color information of the original image for the divided patches. So our results have little borderline.