Depth Map Generation Algorithm for Multiview Video

In this paper, we propose a new approach to depth map generation for multiview video system. Our goal is to improve 3D scene representation consistency by exchanging spatial information between all depth maps in multiview video. The presented algorithm mainly adopts Scale Invariant Feature Transform (SIFT) and jointly disparity and disparity estimation technique. This algorithm first uses the SIFT to get the SIFT descriptors, and then uses Markov random field (MRF) to generate depth map according to the relationship of the SIFT descriptors in different views. This method can reduce computational complexity and adaptively narrow search ranges considering the reliability of each SIFT descriptors. Simulation results show that this algorithm greatly improves the computational accuracy of depth map generation for the multiview video system.