An element image array generation algorithm for enhancing the depth of field quality of 3D reproduction based on multi-depth fusion

• Published in: Optics and lasers in engineering Vol. 195; p. 109282
• Main Authors: Wang, Lu, Wang, Yu, Liu, Quanyang
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.12.2025
• Subjects: Adaptive deep layering,3D reproduction; Element image array; Integral imaging; Integral imaging; Adaptive deep layering,3D reproduction; Element image array
• ISSN: 0143-8166
• DOI: 10.1016/j.optlaseng.2025.109282

•An adaptive hierarchical EIA generation algorithm based on multi-depth fusion is proposed.•By integrating the characteristics of the human visual system and adopting an adaptive Gaussian weight fusion driven by depth difference, the quality of 3D reproduction is improved.•Improve the quality of element image array generation under monocular vision. Monocular-vision-based integral imaging (InIm) offers significant potential for three-dimensional (3D) visualization, enabling naked-eye 3D viewing through a straightforward acquisition process followed by computational imaging. However, the stacking of diffused circles during 3D reconstruction results in a narrow depth-of-field (DOF) range for high-quality display, limiting the widespread adoption of this technology. To address this limitation and enhance display quality, this study presents a multi-depth fusion-based algorithm for generating element image arrays (EIAs). The proposed algorithm leverages the depth information of the 3D scene and display device parameters to construct an adaptive hierarchical model. By incorporating characteristics of the human visual system (HVS) and light field depth cues, it introduces a depth-difference-driven Gaussian fusion coding method. The resulting EIA achieves enhanced 3D reproduction quality within a specified depth range. Simulation and reconstruction experiments were performed on the system's center depth plane (CDP) and two extreme DOF planes. Results demonstrate that the proposed algorithm outperforms comparative methods in the objective metrics of peak signal-to-noise ratio (PSNR) and structural similarity (SSIM), validating its effectiveness.