Depth-based adaptive search range algorithm for motion estimation in HEVC

High efficient video coding has been developed for ultra-high resolution and multi-view videos. It doubles the compression ratio compared to H.264/MPEG-4 AVC, but requires to adopt a very high computational quad-tree structure in motion estimation. Researchers mainly focus on fast mode decision on coding units or prediction units, and have reused the fast motion estimation techniques of H.264/MPEG-4 AVC. However, they do not fully utilize the characteristics of 3D video which is composed of texture streams and depth maps. The depth maps give cues to the objects in the same distance from the projected screen in a 3D scene. In addition to the high temporal correlation between frames, depth maps could be used to link up the objects in consecutive frames such that movements of the same object could be predicted. Therefore, the proposed algorithm aims to define an adaptive search range in motion estimation according to the predicted movements by depth intensity mapping in order to skip the unnecessary search points. Simulation results reveal that the proposed algorithm can reduce the complexity of motion estimation while the coding efficiency can be maintained.