Segmental downsampling intra coding based on spatial locality for HEVC

The new generation standard of video coding, High Efficiency Video Coding (HEVC) achieves better compression performance at the same image quality compared with H.264/AVC. Quad-tree based flexible coding unit structure defines the prediction unit sizes up to 64×64. Plenty of intra prediction modes are employed to achieve better coding efficiency. However, these bring massive calculation and may not perform well enough in high resolution video. This paper proposes a segmental downsampling intra coding scheme which consists of supplemental intra prediction modes, reconstructing reference samples with downsampling and the optimized intra coding process. The proposed supplemental intra prediction modes utilize closer spatial locality reference samples which leads to higher compression efficiency. Optimizing intra coding process reduces the computation complexity and restrains the growth in encoding time. From experimental results, coding efficiency increasing is 18.43% on an average, with only 8.37% encoding time increasing, compared with HEVC.