Simple Baselines for Image Restoration

AbstractAlthough there have been significant advances in the field of image restoration recently, system complexity state-of-the-art (SOTA) methods is increasing as well, which may hinder convenient analysis and comparison methods. In this paper, we propose a simple baseline that exceeds SOTA computationally efficient. To further simplify baseline, reveal nonlinear activation functions, e.g. Sigmoid, ReLU, GELU, Softmax, etc. are not necessary: they could be replaced by multiplication or removed. Thus, derive Nonlinear Activation Free Network, namely NAFNet, from baseline. results achieved on various challenging benchmarks, 33.69 dB PSNR GoPro (for deblurring), exceeding previous 0.38 with only 8.4% its computational costs; 40.30 SIDD denoising), 0.28 less than half costs. The code pre-trained models released at github.com/megvii-research/NAFNet.KeywordsImage restorationImage denoiseImage deblur