mlsanityData to 7 Oct 2026

TM-BSN: Triangular-Masked Blind-Spot Network for Real-World Self-Supervised Image Denoising

arXiv 2604.04484 ↗First posted 6 Apr 2026eess.IV cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

Blind-spot networks (BSNs) enable self-supervised image denoising by preventing access to the target pixel, allowing clean signal estimation without ground-truth supervision. However, this approach assumes pixel-wise noise independence, which is violated in real-world sRGB images due to spatially correlated noise from the camera's image signal processing (ISP) pipeline. While several methods employ downsampling to decorrelate noise, they alter noise statistics and limit the network's ability to utilize full contextual information. In this paper, we propose the Triangular-Masked Blind-Spot Network (TM-BSN), a novel blind-spot architecture that accurately models the spatial correlation of real sRGB noise. This correlation originates from demosaicing, where each pixel is reconstructed from neighboring samples with spatially decaying weights, resulting in a diamond-shaped pattern. To align the receptive field with this geometry, we introduce a triangular-masked convolution that restricts the kernel to its upper-triangular region, creating a diamond-shaped blind spot at the original resolution. This design excludes correlated pixels while fully leveraging uncorrelated context, eliminating the need for downsampling or post-processing. Furthermore, we use knowledge distillation to transfer complementary knowledge from multiple blind-spot predictions into a lightweight U-Net, improving both accuracy and efficiency. Extensive experiments on real-world benchmarks demonstrate that our method achieves state-of-the-art performance, significantly outperforming existing self-supervised approaches. Our code is available at https://github.com/parkjun210/TM-BSN.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 30 of its 46 references that are in the map. First posted on arXiv 6 Apr 2026.

Builds on

30 of 46 references.

References · 30

  1. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,604
  2. 2021Restormer: Efficient Transformer for High-Resolution Image Restorationcited by 1,618
  3. 2016Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoisingcited by 1,352
  4. 2017Understanding the Effective Receptive Field in Deep Convolutional Neural Networkscited by 549
  5. 2018Noise2Noise: Learning Image Restoration without Clean Datacited by 526
  6. 2017FFDNet: Toward a Fast and Flexible Solution for CNN based Image Denoisingcited by 509
  7. 2018Noise2Void - Learning Denoising from Single Noisy Imagescited by 359
  8. 2018Toward Convolutional Blind Denoising of Real Photographscited by 274
  9. 2019Noise2Self: Blind Denoising by Self-Supervisioncited by 274
  10. 2017MemNet: A Persistent Memory Network for Image Restorationcited by 244
  11. 2020CycleISP: Real Image Restoration via Improved Data Synthesiscited by 212
  12. 2017Benchmarking Denoising Algorithms with Real Photographscited by 209
18 more
  1. 2021Neighbor2Neighbor: Self-Supervised Denoising from Single Noisy Imagescited by 171
  2. 2019High-Quality Self-Supervised Deep Image Denoisingcited by 161
  3. 2019Variational Denoising Network: Toward Blind Noise Modeling and Removalcited by 117
  4. 2022AP-BSN: Self-Supervised Denoising for Real-World Images via Asymmetric PD and Blind-Spot Networkcited by 92
  5. 2018Real-world Noisy Image Denoising: A New Benchmarkcited by 74
  6. 2020Unpaired Learning of Deep Image Denoisingcited by 74
  7. 2022C2N: Practical Generative Noise Modeling for Real-World Denoisingcited by 56
  8. 2019When AWGN-based Denoiser Meets Real Noisescited by 53
  9. 2022CVF-SID: Cyclic multi-Variate Function for Self-Supervised Image Denoising by Disentangling Noise from Imagecited by 45
  10. 2023LG-BPN: Local and Global Blind-Patch Network for Self-Supervised Real-World Denoisingcited by 37
  11. 2023Spatially Adaptive Self-Supervised Learning for Real-World Image Denoisingcited by 33
  12. 2023Random Sub-Samples Generation for Self-Supervised Real Image Denoisingcited by 31
  13. 2022Variational Deep Image Restorationcited by 24
  14. 2022Noise2NoiseFlow: Realistic Camera Noise Modeling without Clean Imagescited by 17
  15. 2023Exploring Efficient Asymmetric Blind-Spots for Self-Supervised Denoising in Real-World Scenarioscited by 17
  16. 2023PUCA: Patch-Unshuffle and Channel Attention for Enhanced Self-Supervised Image Denoisingcited by 16
  17. 2023Self-supervised Image Denoising with Downsampled Invariance Loss and Conditional Blind-Spot Networkcited by 12
  18. 2024Rethinking Transformer-Based Blind-Spot Network for Self-Supervised Image Denoisingcited by 9

Built on by

Not cited yet.