search

INVOKERer / DeepRFT

The code for 'Intriguing Findings of Frequency Selection for Image Deblurring' and 'Deep Residual Fourier Transformation for Single Image Deblurring'
Other
223 stars 31 forks source link

border artifacts #26

Open 123456789-qwer opened 2 years ago

123456789-qwer commented 2 years ago

Why does the image output from the test process have border artifacts

INVOKERer commented 2 years ago

Because DeepRFT uses slide window to keep the input size same for test and train.

123456789-qwer commented 2 years ago

How to remove border artifacts? How did you solve it?

INVOKERer commented 2 years ago

You can put the window partition into the fft bench and train the model again, if you want to solve it.

123456789-qwer commented 2 years ago

You can put the window partition into the fft bench and train the model again, if you want to solve it.

Thank you very much for your reply, I would like to know why this can remove artifacts, can you explain the specific principle?

INVOKERer commented 1 year ago

Hi, we solve the border artifacts in https://github.com/DeepMed-Lab-ECNU/DeepRFT-AAAI2023/blob/main/basicsr/models/archs/FMIMOUNet_arch.py by linear fusion.