H.264-based depth map sequence coding using motion information of corresponding texture video

Abstract

Three-dimensional television systems using depth-image-based rendering techniques are attractive in recent years. In those systems, a monoscopic two-dimensional texture video and its associated depth map sequence are transmitted. In order to utilize transmission bandwidth and storage space efficiently, the depth map sequence should be compressed as well as the texture video. Among previous works for depth map sequence coding, H.264 has shown the best performance; however, it has some disadvantages of requiring long encoding time and high encoder cost. In this paper, we propose a new coding structure for depth map coding with H.264 so as to reduce encoding time significantly while maintaining high compression efficiency. Instead of estimating motion vectors directly in the depth map, we generate candidate motion modes by exploiting motion information of the corresponding texture video. Experimental results show that the proposed algorithm reduces the complexity to 60% of the previous scheme that encodes two sequences separately and coding performance is also improved up to 1dB at low bit rates. © 2006 Springer-Verlag.