Depth-texture cooperative clustering and alignment for high efficiency depth intra-coding
Shuai Li, Ce Zhu, Jianjun Lei · 2013
In view of structure similarity between depth and texture in multiview video plus depth, efficient depth intra-coding with the aid of texture information has received a lot of attention. In this paper, a new depth-texture cooperative clustering method is first proposed for cluster-based depth prediction (CBDP) by exploiting the similarity. Due to inaccuracy of depth maps and the resulted texture-depth misalignment along the edges, a small number of residuals after the depth prediction may be of large values, which will greatly compromise the DCT-based coding performance. Accordingly, a simple yet effective detection and rectification scheme is developed to deal with the misalignment problem. The proposed CBDP followed by the misalignment detection and rectification technique is incorporated into the H.264/AVC intra-coding as an additional option for the coding of depth edge blocks. The new depth coding option is shown to achieve rate reduction, while improving SSIM-based quality of the synthesized views.