Spectral analysis of a surface occlusion model for image-based rendering sampling
Weiyan Chen, Changjian Zhu · Digital Signal Processing · 2022
Occlusion is an important and challenging problem in light field rendering technology. We present a signal-processing framework to study the occlusion phenomenon for object surfaces. Our framework combines the occlusion phenomenon with object surface construction information. Based on the geometry of a given scene, we establish an accurate closed-form expression describing object surface occlusion for light field rendering. In this manner, scene occlusion and spectral support of the light field are analyzed in the frequency domain. Light field rendering predictions rooted in the frequency domain then serve as the controlling factors for scene sampling and rendering. This analysis extends previous work on light field sampling theory in which no occlusion in scenes were considered, and it offers a comprehensive treatment of case with self-occlusion and no occlusion. Extensive experimental evaluations of synthetic and actual scenes demonstrate that our light field sampling approach notably outperforms competing algorithms.