Low Complexity Depth Coding Assisted by Coding Information From Color Video

TL;DR: This paper proposes a low complexity depth coding algorithm incorporating a fast motion search method and a fast mode decision method and shows that the proposed algorithm reduces 80% computational complexity with a negligible loss of coding efficiency in comparison with the original Joint Multiview Video Coding encoder.

Abstract: Recently, color plus depth map format 3-D video has been introduced as a new 3-D scene representation format, which gains the attention for its relatively low storage and transmission bandwidth requirement in comparison to the multiview video system. Usually, encoding of the color video and the depth data uses two traditional H.264 encoders. The complexity and hardware requirements to code one 2-D view and its associated depth data are nearly two times higher than coding one 2-D view. In this paper, we address the reduction of complexity of depth coding. Since both color videos and depth maps represent the same scene at the same time instant, it is likely for each macroblock to have similar coding information including the prediction mode and motion vectors. Taking advantage of correlations between coding information from depth maps and color videos, we propose a low complexity depth coding algorithm incorporating a fast motion search method and a fast mode decision method. Results show that the proposed algorithm reduces 80% computational complexity with a negligible loss of coding efficiency in comparison with the original Joint Multiview Video Coding encoder.