Closed XiaotianM closed 6 years ago
We use rf here because when the test image is very large, our model cannot directly process it since the memory is limited. Therefore, we have to split the large image to several patches. The rf means the overlap of the two neighbouring patches. We average the pixel values in the 'rf' region for the final reconstructed image so that to avoid the border effects.
Thanks
In your DRRN test file, you have split the patches from input image because of the memory limit. But I am not sure the usage of 'rf'. You defined as follows:
What's more, It seems that your input patches are more larger than output patches? I am not sure.
can you expalain this for me? Thank you very much.