Multi-occlusion handling in depth estimation of light fields

As the technology of capturing light fields evolves, depth estimation from light field images has become a hot issue in recent years. Nevertheless, how to avoid the estimation error caused by multi-occlusion is still an open-ended problem. In this paper, we present a novel method to handle complex occlusions in depth estimation of light fields. Firstly, we propose a new theory, when a pixel is around the edges and refocused to its correct depth, its spatial patch and angular patch corresponds. Based on this theory, the angular patch is divided into the same regions as the spatial patch. Then, a depth map is improved by computing the cost volumes in multiple angular regions respectively. Different from previous approaches, our proposed method captures sharp depth near the edges in the presence of multi-occlusion. Experimental results demonstrate the effectiveness of the proposed method on multiple synthetic datasets and real world images.