An efficient depth MAP model based on geometric properties

In High Efficiency Video Coding (HEVC) 3D extension depth modeling mode (DMM) is used in depth intra prediction process. According to the original design, DMM contains four modes that are named as DMM1 to DMM4 respectively, and DMM1 contributes obviously coding gain. In DMM1 mode, DMM1 patterns generation is carried out at the initialization process in encoder and decoder side. In this paper, the current patterns generation process will be analyzed, and a new generation mechanism based on geometric properties will be introduced to improve DMM1´s performance and reduce the coding complexity. With help of the new generation mechanism, the number of DMM1 patterns and the encoding time are reduced significantly without coding performance loss.