Texture Mode Dependent Depth Coding in 3D-HEVC.

Xin Zhao,Ying Chen,Li Zhang,Marta Karczewicz
DOI: https://doi.org/10.1109/icip.2013.6738358
2013-01-01
Abstract:In 3D-HEVC, depth modeling modes (DMM) are applied for efficient intra depth coding. With DMM modes, a prediction unit is partitioned into two parts, and each part is predicted by a single value. In one DMM mode, the partition pattern is implicitly derived at the decoder by searching all pre-defined Wedgelet patterns on a Co-located Texture Luma Block (CTLB), which increases the decoding complexity drastically. To simplify the design of this DMM mode, in this paper, we propose to utilize the Intra Prediction Mode (IPM) of CTLB to largely skip unnecessary Wedgelet searches at the decoder. To achieve this goal, for each IPM, only a limited number of Wedgelet patterns are selected as candidates in this DMM mode. Experimental results demonstrate that, with almost no coding performance degradation, the proposed method significantly reduces the decoding complexity by skipping 90% of Wedgelet searches. The proposed method has been adopted by 3D-HEVC.
What problem does this paper attempt to address?