Research On Width First Quad-Tree CU Division And The Algorithm Of Improving Video Coding Efficiency Based On AVS2

AVS2(Audio Video Standard ?)is the new audio video coding standard based on AVS1 which possesses independent Chinese intelectual property rights.Same as H.264 and HEVC(High Efficiency Audio and Video Coding),the goal of this standard is to solve the problem of video data compressing and audio video coding technology is its another name.AVS2 formally promulgated standards for radio and TV industry in May 2016.AVS2 has made great contribution to our national in HD(high definition)and UHD(ultra high definition)and other information industry.Although AVS2 has increased one times coding efficiency compared to the previous AVS1,but it still has relatively high complexity and can not be used in real-time encoding.So,it is of great significance to do some deep research in AVS2.Firstly,This paper introduces the important part intra prediction and the reference software RD14.0.The study found that the quad-tree division for CU(Coding Unit)is time-consuming because of the large amounts modes.This paper provide a new algorithm based on width first for fast CU division after lots of research.The algorithm abandons the original Width First stragey and adopts the Depth First stragy,and then,the algorithm makes use of the difference of RDCost(Rate Distortion Cost)between large CU size and small CU size.So,it can stop the CU division earlier.Because of the early-stop division it can save much more coding time and improve the coding efficiency.The experimental result shows that the algorithm can save nearly 35 percent encoding time with less then 0.1 dB PSNR loss.Then,three algorithms have been studied deeply.The first one is the width-first fast CU division algorithm introduced in chapter 3(algorithm ?).The second one is the fast intra-prediction algorithm based on modes-group(algorithm ?).The third one is the fast CU division algorithm based on spatial correlation(algorithm ?).Analysis found that these three algorithm are in different links in the process of video coding,so they have the possibility of joint application.In this paper,after analysis of these three algorithms,it proves that they could be allianced in the two aspects of theory and experiment and finally the alliance has been completed successfully.Compared to the AVS2,after combining algorithm ? and algorithm ?,the encoding efficiency could be improved 60 percent with negnigible PSNR loss.After combining algorithm ? and algorithm ?,the encoding efficiency could be improved 45 percent with less then 0.1 dB PSNR loss.This paper consists of four chapters.The first chapter introduces the history of the AVS2 development and the current research on it.The second chapter introduces the standard way to divide one CU and the technology of intra-prediction.The third chapter introduces the fast CU division algorithm based on width first and its pinciple.The fourth chapter introduces the union application of several algorithms.