Papers › MaIR: A Locality- and Continuity-Preserving Mamba for Image Restoration

MaIR: A Locality- and Continuity-Preserving Mamba for Image Restoration

28 Dec 2024CVPR 2025 1arXiv:2412.20066archive 2025-07-28

Boyun Li, Haiyu Zhao, Wenxin Wang, Peng Hu, Yuanbiao Gou, Xi Peng

Recent advancements in Mamba have shown promising results in image restoration. These methods typically flatten 2D images into multiple distinct 1D sequences along rows and columns, process each sequence independently using selective scan operation, and recombine them to form the outputs. However, such a paradigm overlooks two vital aspects: i) the local relationships and spatial continuity inherent in natural images, and ii) the discrepancies among sequences unfolded through totally different ways. To overcome the drawbacks, we explore two problems in Mamba-based restoration methods: i) how to design a scanning strategy preserving both locality and continuity while facilitating restoration, and ii) how to aggregate the distinct sequences unfolded in totally different ways. To address these problems, we propose a novel Mamba-based Image Restoration model (MaIR), which consists of Nested S-shaped Scanning strategy (NSS) and Sequence Shuffle Attention block (SSA). Specifically, NSS preserves locality and continuity of the input images through the stripe-based scanning region and the S-shaped scanning path, respectively. SSA aggregates sequences through calculating attention weights within the corresponding channels of different sequences. Thanks to NSS and SSA, MaIR surpasses 40 baselines across 14 challenging datasets, achieving state-of-the-art performance on the tasks of image super-resolution, denoising, deblurring and dehazing. Our codes will be available after acceptance.

PaperPDFConference PDFCode

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

Code

XLearning-SCU/2025-CVPR-MaIR officialmentioned on GitHubpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

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

Tasks

DeblurringDenoisingImage DeblurringImage DehazingImage DenoisingImage RestorationImage Super-ResolutionMambaSingle Image DehazingSuper-Resolution

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Image Deblurring GoPro MaIR PSNR 33.69 #19 of 55 Archive leaderboard report
Image Dehazing SOTS Indoor MaIR PSNR 39.45 #16 of 34 Archive leaderboard report
Image Dehazing SOTS Indoor MaIR SSIM 0.997 #16 of 34 Archive leaderboard report
Image Dehazing SOTS Outdoor MaIR PSNR 36.96 #13 of 31 Archive leaderboard report
Image Dehazing SOTS Outdoor MaIR SSIM 0.991 #13 of 31 Archive leaderboard report
Image Denoising BSD68 sigma50 MaIR+ PSNR 30.08 #1 of 2 Archive leaderboard report
Image Denoising BSD68 sigma50 MaIR PSNR 28.66 #2 of 2 Archive leaderboard report
Image Denoising Urban100 sigma25 MaIR+ PSNR 33.3 #1 of 2 Archive leaderboard report
Image Denoising Urban100 sigma25 MaIR PSNR 33.22 #2 of 2 Archive leaderboard report
Image Denoising Urban100 sigma50 MaIR+ PSNR 30.41 #1 of 4 Archive leaderboard report
Image Denoising Urban100 sigma50 MaIR PSNR 30.3 #2 of 4 Archive leaderboard report
Image Denoising urban100 sigma15 MaIR+ PSNR 35.42 #3 of 4 Archive leaderboard report
Image Denoising urban100 sigma15 MaIR PSNR 35.35 #4 of 4 Archive leaderboard report
Image Super-Resolution Manga109 - 4x upscaling MaIR+ PSNR 32.66 #12 of 50 Archive leaderboard report
Image Super-Resolution Manga109 - 4x upscaling MaIR+ SSIM 0.9297 #12 of 50 Archive leaderboard report
Image Super-Resolution Manga109 - 4x upscaling MaIR PSNR 32.46 #15 of 50 Archive leaderboard report
Image Super-Resolution Manga109 - 4x upscaling MaIR SSIM 0.9284 #15 of 50 Archive leaderboard report
Image Super-Resolution Set14 - 2x upscaling MaIR PSNR 34.75 #11 of 35 Archive leaderboard report
Image Super-Resolution Set14 - 2x upscaling MaIR SSIM 0.9268 #11 of 35 Archive leaderboard report
Image Super-Resolution Set14 - 4x upscaling MaIR+ PSNR 29.28 #15 of 104 Archive leaderboard report
Image Super-Resolution Set14 - 4x upscaling MaIR+ SSIM 0.7974 #15 of 104 Archive leaderboard report
Image Super-Resolution Set14 - 4x upscaling MaIR PSNR 29.2 #21 of 104 Archive leaderboard report
Image Super-Resolution Set14 - 4x upscaling MaIR SSIM 0.7958 #21 of 104 Archive leaderboard report
Image Super-Resolution Set5 - 2x upscaling MaIR+ PSNR 38.62 #12 of 41 Archive leaderboard report
Image Super-Resolution Set5 - 2x upscaling MaIR+ SSIM 0.963 #12 of 41 Archive leaderboard report
Image Super-Resolution Set5 - 2x upscaling MaIR PSNR 38.56 #13 of 41 Archive leaderboard report
Image Super-Resolution Set5 - 2x upscaling MaIR SSIM 0.9628 #13 of 41 Archive leaderboard report
Image Super-Resolution Set5 - 4x upscaling MaIR+ PSNR 33.14 #6 of 12 Archive leaderboard report
Image Super-Resolution Set5 - 4x upscaling MaIR+ SSIM 0.9058 #6 of 12 Archive leaderboard report
Image Super-Resolution Set5 - 4x upscaling MaIR PSNR 32.93 #7 of 12 Archive leaderboard report
Image Super-Resolution Set5 - 4x upscaling MaIR SSIM 0.9045 #7 of 12 Archive leaderboard report
Image Super-Resolution Urban100 - 4x upscaling MaIR+ PSNR 27.89 #13 of 65 Archive leaderboard report
Image Super-Resolution Urban100 - 4x upscaling MaIR+ SSIM 0.8336 #13 of 65 Archive leaderboard report
Image Super-Resolution Urban100 - 4x upscaling MaIR PSNR 27.71 #14 of 65 Archive leaderboard report
Image Super-Resolution Urban100 - 4x upscaling MaIR SSIM 0.8305 #14 of 65 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

AttentionMambaSoftmax

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