Research On Optimization Algorithm Of Intra Prediction In HEVC

Accompanied by the development of Ultra High-Definition(Ultra HD), 4th generation communication system(4G) and mobile devices, video applications got a booming progress. The video mode combines platforms, contents and terminals is gradually becoming mainstream way in the video industry in the future. On considering the increasing video contents and the more important roles of entrances of mobile terminals, higher requests are been brought forward in real-time communication and the power of digital video processing. To meet the demands of growing video applications, Joint Collaborative Team on Video Coding(JCT-VC) is established by MPEG and VCEG which aims at improving the video coding efficiency. A newly High Efficiency Video Coding standard(HEVC) was published by JCT-VC in April 2013. The main goal of the HEVC standardization effort is to enable significantly improved compression performance relative to existing standards H.264/AVC in the range of 50% bit rate reduction and double the compression ratio for equal perceptual video quality. It also supports the Ultra HD with 4k or 8k resolutions and becomes the mainstream video coding standard.HEVC(High Efficient Video Coding) employs the same block-based approach used in previous standards, and re-defined a new syntax of units, called the coding units(CU), the prediction units(PU) and the transform units(TU). A CU can be partitioned into four small sub-CUs with the same size. One CU can only split in four times, and each partitions correspond to different coding depths. 33 directional modes plus DC(flat overall averaging) and Planar(surface fitting) modes can be use in intra-picture prediction on one coding depth. The final optimized prediction mode is selected by thorough search of 35 prediction modes depended on the RD(rate and distortion) cost. Although this method can improve the coding efficiency, it brings high computational complexity and cannot reach the requirements of coding speed in some video applications.This paper presents a fast depth decision method depends on the traditional intra prediction algorithm for reducing unnecessary calculation amount by taking the texture characters into account. This method provides two improvements: texture based fast terminal method by separating blocks into smooth blocks and complex blocks and more complex blocks; relevance based optimized depth prediction method by considering the coding information of adjacent blocks to skip the larger CU.To evaluate the coding performance of the proposed intra prediction, extensive experiences are conducted on several typical video sequences with different QPs. The coding performance and coding complexity is computed and compared with the HEVC reference software HM. Experimental results shows that the proposed fast depth decision method can provides a decreasing of 39.8% in the HEVC intra prediction processing time, with a little increase in the bitrate(1.37% on average).