Occlusion Removal Using Depth Mapping in Three-Dimensional Integral Imaging

In this paper, we present an occlusion removal method to enhance reconstructed images of occluded three-dimensional (3D) objects based on pixel depth mapping technique in 3D multi-perspective imaging. Depth mapping is achieved by minimizing the statistical variance of projection image pixels of 3D object points on different perspective images. Then depth map and variance map are utilized to classify elemental image pixels into object or occlusion classes. Based on pixel classification results, a modified reconstruction algorithm is used to reconstruct the object without the effect of occlusion. In the proposed method, the occlusion is unknown and may be arbitrarily placed in front of the scene. Experimental results are presented to evaluate the effectiveness of the proposed method.