Thanks for your work!
After reading your paper and the code, I've got a question about the output of the DSN. As mentioned in your paper, "The degradation simulation network G^L in our method indeed learns a specific degradation kernel encoded inside the image......". However, the Eq.3 and Fig.3 (a) indicate that the output of DSN is an image. So does DSN actually simulates the whole degradation process and can directly generate degraded images given an input? Or, DSN outputs a kernel, and how do you do with this kernel?
Thanks for your work! After reading your paper and the code, I've got a question about the output of the DSN. As mentioned in your paper, "The degradation simulation network G^L in our method indeed learns a specific degradation kernel encoded inside the image......". However, the Eq.3 and Fig.3 (a) indicate that the output of DSN is an image. So does DSN actually simulates the whole degradation process and can directly generate degraded images given an input? Or, DSN outputs a kernel, and how do you do with this kernel?