search

wyhuai / DDNM

[ICLR 2023 Oral] Zero-Shot Image Restoration Using Denoising Diffusion Null-Space Model
MIT License
1.08k stars 75 forks source link

How to direct SR on images without preliminary degradation? #70

Open rabbitlingli opened 2 months ago

rabbitlingli commented 2 months ago

Dear Author,

Greetings to you. Firstly, please accept my heartfelt appreciation for kindly providing such excellent code. Moving forward, I have a query that emerged while perusing both the readme and the evaluation.sh script. It appears that prior to conducting any super-resolution (SR) work, it is mandatory to apply either a "degradation operator" or perform a "deg_scale" operation. May I directly employ the SR algorithm on my images without resorting to the "degradation operator"? If so, could you kindly guide me on how to proceed accordingly?

Your expert guidance in this matter would be immensely appreciated.

Thank you in advance for your time and assistance.

Best regards lilingli

wyhuai commented 2 months ago

Hi, our method needs to specify the degradation operator. You can try methods like GFP-GAN which do without defining the degradation operator.