Improving Image Restoration through Removing Degradations in Textual Representations

Open paper
Benchmark Model Rank Results
color-image-denoising-on-cbsd68-color-gaussian-denoising-50TextualDegRemoval + Restormer (Gaussian denoising, σ=50) (extra data)#1PSNR: 28.83
color-image-denoising-on-cbsd68-color-gaussian-denoising-15TextualDegRemoval + Restormer (Gaussian denoising, σ=15) (extra data)#2PSNR: 34.48
color-image-denoising-on-cbsd68-color-gaussian-denoising-25TextualDegRemoval + Restormer (Gaussian denoising, σ=25) (extra data)#2PSNR: 31.97
image-denoising-on-siddTextualDegRemoval + Restormer (SIDD) (extra data)#4PSNR (sRGB): 40.09SSIM (sRGB): 0.960
image-deblurring-on-goproTextualDegRemoval + NAFNet (trained on GoPro) (extra data)#11PSNR: 33.97SSIM: 0.968
image-dehazing-on-sots-indoorTextualDegRemoval + SFNet (SOTS Indoor) (extra data)#13PSNR: 41.48SSIM: 0.996

Rank counts only results with a code link.