Edge-oriented Depth Map Coding
Video plus depth representation has been presented as an efficient format for threedimensional (3D) video. Depth image-based rendering (DIBR) enables the generation of virtual view for multi-view 3D display and free viewpoint video. By analyzing the effect of distortion caused by geometry error of coded depth map, this paper proposes an adaptive edge-oriented method for depth map coding, which combines filter and different compression methods for different regions. In proposed method, edge in depth image is extracted and masks are applied for segmentation. Then lossless Huffman compression is performed to extract edge area, for preserving the edge information, and an embedded zero tree wavelet (EZW) based algorithm is used to encode other areas with different coefficients. The vertical edge can also be used for application of an asymmetry filter to avoid artifacts. Experimental results show the proposed method saves bitrates of depth map and improves the encoding performance, with coding gain of 0.1-0.8 db in synthesized frame quality.
multi-view plus depth DIBR depth map coding 3d video
Jing Guo Feng Yang Liang Qian
School of Electronic. Information and Electrical Engineering Shanghai Jiao Tong University Shanghai, School of Electronic, Information and Electrical Engineering Shanghai Jiao Tong University Shanghai,
国际会议
海口
英文
379-382
2011-07-15（万方平台首次上网日期，不代表论文的发表时间）