Spatial locality based supplemental modes for intra prediction of HEVC

The latest generation standard of video coding, High Efficiency Video Coding (HEVC) achieves the goal of requiring 50% bit-rate of the H.264/AVC at the same image quality. It adopts quad-tree based flexible coding unit structure as prediction unit (PU) sizes which are defined from 64×64 to 4×4, and plenty of intra prediction modes whose reference samples are spatially reconstructed neighboring samples which are at the bottom left, left, top left, above, above right. However, such modes may not come off well enough in high resolution video compression. This paper presents a novel prediction algorithm for intra prediction of HEVC. The proposed supplemental modes based on spatial locality can utilize the spatially neighboring referencing samples more efficient. Compared with the original HEVC, the result of coding efficiency increasing is about 8.04%.