Questions about resize quality?

[TheWangYang]

Why do I use the first weight "EDSR-baseline-SRNO" provided in this repository to resize my own picture? The effect of using the resize method integrated in opencv or pillow is not as good?

Looking forward to your reply, I will be very grateful.

[2y7c3]

Could you provide more details？

[TheWangYang]

@2y7c3 Thank you very much for your prompt reply!

I am very sorry that I was in a hurry to ask questions yesterday and caused confusion to you. My specific questions are:

I now have a task that needs to enlarge some pictures of different sizes (called data set A) to a fixed size. I first used the resize() method in the pillow library and the opencv library.

Then, I use your work (using the weight "EDSR-baseline-SRNO") to first enlarge the data set A by 2 times to get the processed data set B, and then use the resize() method in the opencv or pillow library Based on the data set B, the second step is scaled to the size required by the task.

The PSNR result of the picture I obtained in this way is lower than the result obtained by directly using the resize() method in the opencv or pillow library on the data set A. I would like to consult, what is the main reason for this problem?