Exploring Efficient Asymmetric Blind-Spots for Self-Supervised Denoising in Real-World Scenarios

Abstract

Self-supervised denoising has attracted widespread at-tention due to its ability to train without clean images. How-ever, noise in real-world scenarios is often spatially cor-related, which causes many self-supervised algorithms that assume pixel-wise independent noise to perform poorly. Re-cent works have attempted to break noise correlation with downsampling or neighborhood masking. However, denoising on downsampled subgraphs can lead to aliasing effects and loss of details due to a lower sampling rate. Further-more, the neighborhood masking methods either come with high computational complexity or do not consider local spatial preservation during inference. Through the analy-sis of existing methods, we point out that the key to obtaining high-quality and texture-rich results in real-world self-supervised denoising tasks is to train at the original input resolution structure and use asymmetric operations during training and inference. Based on this, we propose Asymmet-ric Tunable Blind-Spot Network (AT-BSN), where the blind-spot size can be freely adjusted, thus better balancing noise correlation suppression and image local spatial destruction during training and inference. In addition, we regard the pre-trained AT-BSN as a meta-teacher network capable of generating various teacher networks by sampling different blind-spots. We propose a blind-spot based multi-teacher distillation strategy to distill a lightweight network, signif-icantly improving performance. Experimental results on multiple datasets prove that our method achieves state-of-the-art, and is superior to other self-supervised algorithms in terms of computational overhead and visual effects.