Papers › SwinIA: Self-Supervised Blind-Spot Image Denoising without Convolutions

SwinIA: Self-Supervised Blind-Spot Image Denoising without Convolutions

9 May 2023arXiv:2305.05651archive 2025-07-28

Mikhail Papkov, Pavel Chizhov, Leopold Parts

Self-supervised image denoising implies restoring the signal from a noisy image without access to the ground truth. State-of-the-art solutions for this task rely on predicting masked pixels with a fully-convolutional neural network. This most often requires multiple forward passes, information about the noise model, or intricate regularization functions. In this paper, we propose a Swin Transformer-based Image Autoencoder (SwinIA), the first fully-transformer architecture for self-supervised denoising. The flexibility of the attention mechanism helps to fulfill the blind-spot property that convolutional counterparts normally approximate. SwinIA can be trained end-to-end with a simple mean squared error loss without masking and does not require any prior knowledge about clean data or noise distribution. Simple to use, SwinIA establishes the state of the art on several common benchmarks.

PaperPDF

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

No code repository is listed for this paper in the archive or in Syntology's graph.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Color Image DenoisingDenoisingGrayscale Image DenoisingImage DenoisingMedical Image Denoising

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Color Image Denoising BSD300 lambda30 SwinIA PSNR 27.92 #1 of 1 Archive leaderboard report
Color Image Denoising BSD300 lambda30 SwinIA SSIM 0.775 #1 of 1 Archive leaderboard report
Color Image Denoising BSD300 lambda5-50 SwinIA PSNR 27.74 #1 of 1 Archive leaderboard report
Color Image Denoising BSD300 lambda5-50 SwinIA SSIM 0.764 #1 of 1 Archive leaderboard report
Color Image Denoising BSD300 sigma25 SwinIA PSNR 28.4 #1 of 1 Archive leaderboard report
Color Image Denoising BSD300 sigma25 SwinIA SSIM 0.789 #1 of 1 Archive leaderboard report
Color Image Denoising BSD300 sigma5-50 SwinIA PSNR 28.4 #1 of 1 Archive leaderboard report
Color Image Denoising BSD300 sigma5-50 SwinIA SSIM 0.785 #1 of 1 Archive leaderboard report
Color Image Denoising Kodak24 lambda30 SwinIA PSNR 29.51 #1 of 1 Archive leaderboard report
Color Image Denoising Kodak24 lambda30 SwinIA SSIM 0.805 #1 of 1 Archive leaderboard report
Color Image Denoising Kodak24 lambda5-50 SwinIA PSNR 29.06 #1 of 1 Archive leaderboard report
Color Image Denoising Kodak24 lambda5-50 SwinIA SSIM 0.788 #1 of 1 Archive leaderboard report
Color Image Denoising Kodak24 sigma25 SwinIA PSNR 30.12 #3 of 3 Archive leaderboard report
Color Image Denoising Kodak24 sigma25 SwinIA SSIM 0.819 #3 of 3 Archive leaderboard report
Color Image Denoising Kodak24 sigma5-50 SwinIA PSNR 30.3 #1 of 1 Archive leaderboard report
Color Image Denoising Kodak24 sigma5-50 SwinIA SSIM 0.820 #1 of 1 Archive leaderboard report
Color Image Denoising Set14 lambda30 SwinIA PSNR 28.74 #1 of 1 Archive leaderboard report
Color Image Denoising Set14 lambda30 SwinIA SSIM 0.799 #1 of 1 Archive leaderboard report
Color Image Denoising Set14 lambda5-50 SwinIA PSNR 28.27 #1 of 1 Archive leaderboard report
Color Image Denoising Set14 lambda5-50 SwinIA SSIM 0.780 #1 of 1 Archive leaderboard report
Color Image Denoising Set14 sigma25 SwinIA PSNR 29.54 #1 of 1 Archive leaderboard report
Color Image Denoising Set14 sigma25 SwinIA SSIM 0.814 #1 of 1 Archive leaderboard report
Color Image Denoising Set14 sigma5-50 SwinIA PSNR 29.49 #1 of 1 Archive leaderboard report
Color Image Denoising Set14 sigma5-50 SwinIA SSIM 0.809 #1 of 1 Archive leaderboard report
Grayscale Image Denoising BSD68 sigma15 SwinIA PSNR 31.07 #16 of 16 Archive leaderboard report
Grayscale Image Denoising BSD68 sigma15 SwinIA SSIM 0.856 #16 of 16 Archive leaderboard report
Grayscale Image Denoising BSD68 sigma25 SwinIA PSNR 29.17 #12 of 16 Archive leaderboard report
Grayscale Image Denoising BSD68 sigma25 SwinIA SSIM 0.801 #12 of 16 Archive leaderboard report
Grayscale Image Denoising BSD68 sigma50 SwinIA PSNR 26.61 #3 of 15 Archive leaderboard report
Grayscale Image Denoising BSD68 sigma50 SwinIA SSIM 0.706 #3 of 15 Archive leaderboard report
Grayscale Image Denoising Hanzi SwinIA PSNR 14.35 #1 of 1 Archive leaderboard report
Grayscale Image Denoising Hanzi SwinIA SSIM 0.556 #1 of 1 Archive leaderboard report
Grayscale Image Denoising Set12 sigma15 SwinIA PSNR 30.37 #7 of 8 Archive leaderboard report
Grayscale Image Denoising Set12 sigma15 SwinIA SSIM 0.857 #7 of 8 Archive leaderboard report
Grayscale Image Denoising Set12 sigma25 SwinIA PSNR 28.72 #6 of 6 Archive leaderboard report
Grayscale Image Denoising Set12 sigma25 SwinIA SSIM 0.817 #6 of 6 Archive leaderboard report
Grayscale Image Denoising Set12 sigma50 SwinIA PSNR 26.03 #8 of 8 Archive leaderboard report
Grayscale Image Denoising Set12 sigma50 SwinIA SSIM 0.736 #8 of 8 Archive leaderboard report
Medical Image Denoising FMD Confocal Fish SwinIA PSNR 31.79 #1 of 1 Archive leaderboard report
Medical Image Denoising FMD Confocal Fish SwinIA SSIM 0.871 #1 of 1 Archive leaderboard report
Medical Image Denoising FMD Confocal Mice SwinIA PSNR 37.65 #1 of 1 Archive leaderboard report
Medical Image Denoising FMD Confocal Mice SwinIA SSIM 0.960 #1 of 1 Archive leaderboard report
Medical Image Denoising FMD Two-Photon Mice SwinIA PSNR 33.25 #1 of 1 Archive leaderboard report
Medical Image Denoising FMD Two-Photon Mice SwinIA SSIM 0.915 #1 of 1 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

AttentionDense ConnectionsLayer NormalizationLinear LayerMulti-Head AttentionRelative Position EncodingsResidual ConnectionSoftmaxSwin Transformer

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections