# VSANet: View-aware Sparse Attention Network for Light Field Image Denoising

arXiv: https://arxiv.org/abs/2606.24737 · HTML: https://mlsanity.com/paper/2606.24737/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 10 of its 24 references that are in the map.
- First posted on arXiv 23 Jun 2026.

## Builds on (most cited first, 10 of 10)

- [CBAM: Convolutional Block Attention Module](https://mlsanity.com/paper/1807.06521/) (2018)
- [Reformer: The Efficient Transformer](https://mlsanity.com/paper/2001.04451/) (2020)
- [Light Field Image Super-Resolution with Transformers](https://mlsanity.com/paper/2108.07597/) (2021)
- [A 4D Light-Field Dataset and CNN Architectures for Material Recognition](https://mlsanity.com/paper/1608.06985/) (2016)
- [Accurate Light Field Depth Estimation with Superpixel Regularization over Partially Occluded Regions](https://mlsanity.com/paper/1708.01964/) (2017)
- [Selective Light Field Refocusing for Camera Arrays Using Bokeh Rendering and Superresolution](https://mlsanity.com/paper/2108.03918/) (2021)
- [Light Field Denoising via Anisotropic Parallax Analysis in a CNN Framework](https://mlsanity.com/paper/1805.12358/) (2018)
- [Probabilistic-based Feature Embedding of 4-D Light Fields for Compressive Imaging and Denoising](https://mlsanity.com/paper/2306.08836/) (2023)
- [Beyond Subspace Isolation: Many-to-Many Transformer for Light Field Image Super-resolution](https://mlsanity.com/paper/2401.00740/) (2024)
- [LGFN: Lightweight Light Field Image Super-Resolution using Local Convolution Modulation and Global Attention Feature Extraction](https://mlsanity.com/paper/2409.17759/) (2024)

## Abstract (arXiv)

Light field (LF) image denoising is challenging due to the high-dimensional structure of LF data. While noise is independent across sub-aperture images, scene content exhibits strong cross-view correlations. We introduce VSANet, a view-aware sparse attention network for LF denoising. Specifically, we propose a view-aware sparse attention (VSA) block that represents the 4D LF feature map as a unified spatial-angular token space and performs cross-view aggregation via locality-sensitive hashing-based sparse attention. This enables global feature interactions with linear complexity, effectively exploiting LF correlations across views and spatial locations. In addition, we design a feature refinement (FR) block to emphasize informative features in spatial, angular, and epipolar subspaces. The VSA and FR blocks are integrated within a sequential attention refinement module, forming the core of VSANet. Experiments demonstrate VSANet outperforms stateof-the-art LF denoising methods.