The Acceleration Algorithms Of HEVC Intra Coding

Video compressing technology is a precondition for video processing, because original video contains a huge mount of data information, it was a challenge for storage capacity and computing capacity. Video compressing technology utilizes the redundant information in video content, and reduces the bit number of coding image without im-age quality loss. The newest video compressing standard is the HEVC (High Efficiency Video Coding), which published by JCT-VC in 2013. Experimental result shows that, compared to previous standard, HEVC can save about 50% bit-rate on average with same image quality. The new technology improves the coding efficiency while bring-ing more complicated calculation, it limits the application and popularization of HEVC greatly. Especially in some coding areas which need real-time, such as surveillance video, portable mobile devices and video conferencing. So it is meaningful to study fast coding algorithm for HEVC. Intra coding plays an important part in HEVC tech-nology, it brings a quad-tree coding structure. Firstly, a frame is divided into some CTUs(Coding Tree Unit) and each CTU can be divided to smaller CUs(Coding Unit) recursively, CU size ranges from 64×64 to 8×8. Secondly, the algorithm traverses all the CUs to derive a best CU partition. This paper study fast algorithms of intra coding mainly, the detailed contents as follows:1. Fast coding algorithm based on GLCM(Gray Level Co-occurrence Matrix):A large number of studies show that, the CU size in intra coding has strong relevance with image content. In smooth area, HEVC encoder always use larger CU size, while in the area with lush content and image details, HEVC encoder always use smaller CU size such as 16*16 and 8*8. We analyze image texture complexity by GLCM to accelerate CU partition processing.2. Fast coding algorithm based on RD(Rate Distortion) cost:Through analysis, we found that the derived RD cost has strong relevance with CU partition, We can utilize spatial locality of image and analyze RD cost distribution under different encoding size. In addition, for most videos, there are only little part of regions changed in consecutive frames, so we can determine the partitioned depth range of current CTU using the en-coded part in previous frame. Meanwhile, we also use the distribution of histogram to determine the partitioned depth.For verifying the proposed algorithms, we implement these methods on HM13.0. Experimental results shows that, the proposed algorithms can significantly saving en-coding time with negligible image loss and reduce the coding complexity.