A Benchmark for Heterogeneous Stereo Deblurring with Physically- and Epipolar-constrained Cross Attention

2026-06-24

AI Summary

This work addresses the challenge of asymmetric blur in heterogeneous stereo cameras—such as those in smartphones—where hardware discrepancies violate the homogeneity assumptions underlying most existing deblurring methods. To tackle this issue, the authors introduce the first real-world heterogeneous stereo deblurring (HSD) dataset captured using actual mobile devices and propose a lightweight Physical and Epipolar Cross-Attention (PECA) module. PECA integrates an optically derived disparity upper bound with epipolar geometric constraints to enable reliable cross-view feature matching and fusion within a physically plausible disparity range. It further incorporates confidence-weighted aggregation and an adaptive fallback mechanism. Experiments demonstrate that models equipped with PECA significantly outperform state-of-the-art approaches on the HSD dataset, achieving superior restoration quality and computational efficiency while maintaining architectural generality.

Abstract

Modern stereo-capable smartphones enable immersive XR content capture. However, hardware heterogeneity across camera modules often causes severe asymmetric blur artifacts. Existing methods and benchmarks largely assume homogeneous stereo setups and therefore do not explicitly address such asymmetric degradation. To bridge this gap, we present a dedicated framework for heterogeneous stereo deblurring. First, we introduce the heterogeneous stereo deblurring (HSD) dataset, constructed from real smartphone stereo captures via multi-frame integration. Second, we propose physically- and epipolar-constrained cross attention (PECA), a lightweight module that restricts cross-view matching to an epipolar search window bounded by a optics-derived disparity upper bound. By enforcing physically valid disparity constraints, PECA enables efficient and reliable cross-view feature fusion. Moreover, our confidence-weighted attention with residual fusion emphasizes cross-guided deblurring when correspondences are reliable, while naturally falling back to self-deblurring in occluded or unreliable regions. PECA is architecture-agnostic and consistently improves CNN-, Transformer-, and NAFNet-based baselines. Extensive experiments on HSD show that PECA-enhanced models achieve improved restoration performance with favorable efficiency.

Problem

• heterogeneous stereo
• asymmetric blur
• stereo deblurring
• camera hardware heterogeneity
• cross-view degradation

Innovation

• heterogeneous stereo deblurring
• epipolar-constrained cross attention
• physically-constrained disparity
• confidence-weighted attention
• real-world stereo benchmark