Abstract: The rapid developments of 3D video techniques and real applications have attracted the attention of consumers, so the demand degree of storage and transformation for 3D video content has become stronger than before. The 3D-HEVC compression structure is the development of HEVC. 3D-HEVC adopts a multi-view video plus depth (MVD) system, where it establishes correspondences between every depth map and every texture frame. Depth modeling modes (DMM) are also developed by 3D-HEVC because they play an important role in sharp edge encoding. However, the distinct trouble of using DMM, which increases computational burdens, will be introduced in the 3D-HEVC encoder. This paper presents a fast mode selection system that researches the texture frames and depth maps in order to attain the edge data of prediction units (PUs), and then a novel early stop according to the flatness of the current block is also adopted. Beyond that, we also introduce a fresh heuristic method in terms of shuffled frog leaping to obtain the best wedgelet pattern in 3D-HEVC. The overall experiment demonstrates that our algorithm surpasses HTM16.0 in terms of coding efficiency with negligible loss.

Key words: 3D-HEVC, depth, MVD, DMM, SFLA