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

arXiv:2303.16783v2 Announce Type: replace
Abstract: Self-supervised denoising has attracted widespread attention due to its ability to train without clean images. However, noise in real-world scenarios is often spatially correlated, which causes many self-supervised algorithms that assume pixel-wise independent noise to perform poorly. Recent 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. Furthermore, the neighborhood masking methods …

abstract algorithms arxiv attention blind correlation cs.cv denoising however images independent noise pixel train type wise world