Abstract

In this paper, we present an improved 3D reconstruction method of a partially occluded object by using group-based sparse representation (GSR) in integral imaging. In the proposed method, a 3D scene is recorded as elemental images through a camera array. Then, the foreground occlusions are identified and removed based on variance estimation on the sequence of elemental images. Finally, a GSR is used to restore each elemental image which can improve the visual quality of the reconstructed 3D images. The experimental results can confirm the feasibility of the proposed method.

© 2017 Optical Society of America