Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Noise2Self: Blind Denoising by Self-Supervision

About

We propose a general framework for denoising high-dimensional measurements which requires no prior on the signal, no estimate of the noise, and no clean training data. The only assumption is that the noise exhibits statistical independence across different dimensions of the measurement, while the true signal exhibits some correlation. For a broad class of functions ("$\mathcal{J}$-invariant"), it is then possible to estimate the performance of a denoiser from noisy data alone. This allows us to calibrate $\mathcal{J}$-invariant versions of any parameterised denoising algorithm, from the single hyperparameter of a median filter to the millions of weights of a deep neural network. We demonstrate this on natural image and microscopy data, where we exploit noise independence between pixels, and on single-cell gene expression data, where we exploit independence between detections of individual molecules. This framework generalizes recent work on training neural nets from noisy images and on cross-validation for matrix factorization.

Joshua Batson, Loic Royer• 2019

Related benchmarks

TaskDatasetResultRank
Image DenoisingSIDD (val)
PSNR30.72
168
Ultrasound Image DenoisingIn vivo Ultrasound 2 sub-apertures
gCNR91
144
Image DenoisingDND
PSNR33.63
135
Image DenoisingSet12 (test)
PSNR28.37
107
Image DenoisingSIDD (test)
PSNR30.98
102
Image DenoisingCBSD68 (test)
PSNR33.18
100
Image DenoisingSIDD Benchmark
PSNR32.57
97
Gaussian DenoisingBSD68
PSNR28.28
89
Ultrasound Image DenoisingIn vivo Ultrasound 8 sub-apertures
gCNR0.95
72
Ultrasound Image DenoisingIn vivo Ultrasound 8 sub-apertures
CNR9.04
72
Showing 10 of 94 rows
...

Other info

Follow for update