High-Quality Self-Supervised Deep Image Denoising

By Samuli Laine, Tero Karras, Jaakko Lehtinen, Timo Aila. Neural Information Processing Systems, 2019

424 citations recorded by the source corpus.

We describe a novel method for training high-quality image denoising models based on unorganized collections of corrupted images. The training does not need access to clean reference images, or explicit pairs of corrupted images, and can thus be applied in situations where such data is unacceptably expensive or impossible to acquire. We build on a recent technique that removes the need for reference data by employing networks with a "blind spot" in the receptive field, and significantly improve two key aspects: image quality and training efficiency. Our result quality is on par with state-of-the-art neural network denoisers in the case of i.i.d. additive Gaussian noise, and not far behind with Poisson and impulse noise. We also successfully handle cases where parameters of the noise model are variable and/or unknown in both training and evaluation data.

View Paper: https://www.semanticscholar.org/paper/43eff313f8c9ced25f553669152ed04d5b243cab

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Self2Self With Dropout: Learning Self-Supervised Denoising From Single Image (2020, methodology): “Instead of using organized noisy image pairs, some approaches [15, 16, 3, 17] use only unorganized noisy images for NN training, which is done by deﬁning an effective self-prediction loss.”
Neighbor2Neighbor: Self-Supervised Denoising from Single Noisy Images (2021, methodology): “Recently, Probabilistic Noise2Void [14], Laine19 [15], and Dilated Blind-Spot Network [30] further introduced explicit noise modeling and probabilistic inference for better performance.”
NeRF in the Dark: High Dynamic Range View Synthesis from Noisy Raw Images (2021, methodology): “Various followups to Noise2Noise proposed modified network architectures allowing supervision with a dataset of single noisy images [4, 30, 31].”
Self-Supervised Learning of Graph Neural Networks: A Unified Review (2021, background): “Recently, SSL has shown its promising capability in data restoration tasks, such as image super-resolution [1], image denoising [2, 3, 4], and single-cell analysis [5].”
VIME: Extending the Success of Self- and Semi-supervised Learning to Tabular Domain (2020, background): “Most existing works with pretext tasks are appropriate only for images or natural language: (i) surrogate classes prediction (scaling and translation) [11], (ii) rotation degree predictions [7], (iii) colorization [9], (iv) relative position of patches estimation [12], (v) jigsaw puzzle solving [8],”
BockNet: Blind-Block Reconstruction Network With a Guard Window for Hyperspectral Anomaly Detection (2023, background): “We emphasize that the blind-spot network cannot see the single-pixel information in the center (blind-spot) of the receptive field, but only predicts the blind-spot using its neighboring pixels [67], [68].”