HDR Video Reconstruction with Tri-Exposure Quad-Bayer Sensors

• DOI: 10.48550/arxiv.2103.10982
• Publication Date: 2021-01-01
• Authors: Yitong Jiang; Inchang Choi; Jun Jiang; Jinwei Gu
• Keywords: Motion blur

We propose a novel high dynamic range (HDR) video reconstruction method with new tri-exposure quad-bayer sensors. Thanks to the larger number of exposure sets and their spatially uniform deployment over frame, they are more robust noise spatial artifacts than previous varying (SVE) HDR methods. Nonetheless, motion blur from longer exposures, short inherent SVE methods remain huge obstacles. Additionally, temporal coherence must be taken into account for stability reconstruction. To tackle these challenges, we introduce network architecture that divides-and-conquers problems. In order better adapt large range, also LDR-reconstruction loss takes equal contributions both highlighted shaded pixels frames. Through series comparisons ablation studies, show our solution is optimal capture methods, particularly scenes objects motion.