Lightweight Quad Bayer HybridEVS Demosaicing via State Space Augmented Cross-Attention

Shiyang Zhou, Haijin Zeng, Yongyong Chen, Jie Liu, Jingyong Su

Keywords: Demosaicing, HybridEVS, Quad Bayer, State Space, Attention

Abstract: Event cameras like the Hybrid Event-based Vision Sensor (HybridEVS) camera capture brightness changes as asynchronous "events" instead of frames, offering advantages over traditional cameras: high temporal resolution, wide dynamic range, and no motion blur. However, challenges arise from combining a Quad Bayer Color Filter Array (CFA) sensor with event pixels lacking color information, resulting in aliasing and artifacts on the demosaicing process before downstream application. Current methods struggle to address these issues, especially on resource-limited mobile devices. In response, we introduce \textbf{TSANet}, a lightweight \textbf{T}wo-stage network via \textbf{S}tate space augmented cross-\textbf{A}ttention, which can handle event pixels inpainting and Quad Bayer demosaicing separately, leveraging the benefits of dividing complex tasks into manageable subtasks and learning them through a two-step training strategy to enhance robustness. Additionally, we propose a lightweight Cross-Swin State Block (CSSB) designed to augment the model's capacity to capture global dependencies using state space models in a linear format, along with cross-modality Swin attention to integrate additional priors like CFA pattern and event map, outperforming traditional local attention mechanisms while also reducing model size. In summary, TSANet demonstrates excellent demosaicing performance on HybridEVS while maintaining a lightweight model, averaging better results than the previous state-of-the-art method DemosaicFormer across seven diverse datasets in both PSNR and SSIM, while respectively reducing parameter and computation costs by $1.86\times$ and $3.29\times$. Our approach presents new possibilities for efficient image demosaicing on mobile devices. \textit{Code and models are available in supplementary materials.}

Primary Area: applications to computer vision, audio, language, and other modalities

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 9614