Pseudo- and Reordered Sub-Aperture Image Formats for Light Field Residual Compression

Light field (LF) is a 3D imaging approach that renders photorealistic views from light rays passing through free space, but its large data volume motivates active research in LF compression. This paper proposes three novel LF formats: pseudo sub-aperture image (P-SAI), reordered SAI (RoSAI), and P-RoSAI to improve LF residual compression in synthesis-based LF compression. P-SAI reconstructs SAIs using arbitrary pseudo-viewpoints instead of the fixed viewpoints of the original LF sample, exploiting both spatial and temporal locality while ensuring the theoretical rules of LF are preserved. RoSAI extends the reordering concept into the temporal domain, enabling full utilization of video codec capabilities. P-RoSAI integrates both approaches, leveraging the benefits of P-SAI and RoSAI simultaneously. Experiments across multiple datasets, synthesis difficulty levels, and quantization parameters demonstrate that the proposed formats improve compression efficiency and provide flexible options for future LF compression research.