About noise about dcls-sr

[Algolzw]

As the same as previous works such as KernelGAN, IKC, and DAN, we didn't add noises in the anisotropic setting. However, if you want to evaluate our method on noisy images, you should modify the setting (in the options file) as follows and re-train the model:

degradation:
  random_kernel: True
  ksize: 21
  code_length: 10
  sig_min: 0.2
  sig_max: 4.0
  rate_iso: 1.0
  random_disturb: false
  ######### add noises ###########
  noise: True
  # noise_high: 0.0588 # noise level 15: 15 / 255.
  noise_high: 0.1176 # noise level 30: 30 / 255.
  ################################

[xiab3369]

[xiab3369]

It's different from what you describe in the paper?

[Algolzw]

@xiab3369 Thanks for pointing it out! This is my writing mistake that the noises are only added to the isotropic setting as shown in Table 2. I will be happy to revise it in a new ArXiv version.

And if you want to try some additional experiments on noises, you could just modify the options file as I mentioned above. Thanks again and good luck!