Awesome-Denoise
There are three main factors to divide these papers into different catrgories to have a better idea.
Sometimes raw domain denoising papers would use some ISP to convert to sRGB domain, So use Both to cover this situation.
Sometimes video denoising papers degrade to burst denoising, even single image denoising, always use Video tag to cover this situation.
-
Color Space
- RGB
- Raw
- Both
-
Image Kind
- Single
- Burst
- Video
-
Noise Model
- AWGN(Additive White Gaussian Noise model)
- PG(Posion Gaussian noise model)
- GAN(Gan based noise model)
- Real(camera or dlsr devices real noise model)
- Prior
- Low Rank
- Sparsity
- self similarity
benchmark dataset
- SIDD, CVPR 2018, citation 256
- RENOIR, JVCIR 2018, citation 106
- PolyU, arxiv 2018, citation 108
- SID, CVPR 2018, citation 595
- DND, CVPR 2017, citation 296
- NaM, CVPR 2016, citation 148
self-supervised denoising
video denoising
- Unsupervised deep video denoising
- ICCV 2021, UDVD
- Recurrent Self-Supervised Video Denoising with Denser
Receptive Field
- ACM MM 2023, code
image denoising