12 January 2018 Depth estimation of complex geometry scenes from light fields
Author Affiliations +
The surface camera (SCam) of light fields gathers angular sample rays passing through a 3D point. The consistency of SCams is evaluated to estimate the depth map of scene. But the consistency is affected by several limitations such as occlusions or non-Lambertian surfaces. To solve those limitations, the SCam is partitioned into two segments that one of them could satisfy the consistency constraint. The segmentation pattern of SCam is highly related to the texture of spatial patch, so we enforce a mask matching to describe the shape correlation between segments of SCam and spatial patch. To further address the ambiguity in textureless region, a global method with pixel-wise plane label is presented. Plane label inference at each pixel can recover not only depth value but also local geometry structure, that is suitable for light fields with sub-pixel disparities and continuous view variation. Our method is evaluated on public light field datasets and outperforms the state-of-the-art.
© (2018) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Lipeng Si, Lipeng Si, Qing Wang, Qing Wang, } "Depth estimation of complex geometry scenes from light fields", Proc. SPIE 10620, 2017 International Conference on Optical Instruments and Technology: Optoelectronic Imaging/Spectroscopy and Signal Processing Technology, 106200M (12 January 2018); doi: 10.1117/12.2291596; https://doi.org/10.1117/12.2291596

Back to Top