The Research On Fast Video Coding Algorithm For HEVC

HEVC is the newest video coding standard developed by JCT-VC organization,which is a joint collaboration team between VCEG and MPEG.Compared to the previous generation video coding standard,some new techniques are introduced in HEVC to make higher compression ratio and better image quality.However,these new techniques brings high computational complexity.How to reduce the complexity while ensuring coding performance is critical issue for HEVC to be widely used.To solve the problem of long coding time and high computational complexity of HEVC,the new innovation focused on intra and inter coding are carried out in this dissertion.The main contributions are summarized as follows:Firstly,intra prediction is key technology to eliminate the spatial redundancy in HEVC encoder,Howerever the complicated process of quadtree-based coding unit blocking partition and mode selection add high computational complexity for intra coding.Under consideration of reducing computational complexity,the improvement of intra prediction is made in this paper.Firstly the spatio-temporal correlation of adjacent coding unit depth is used to narrow the depth range and then unnecessary depth can be split or prune early.At the same time,according to whether Planar is selected as the optimal prediction mode,the number of prediction modes is reduced in different methods.Compared with original HM10.0,experimental results show that the proposed algorithm can save nearly 30%encoding time with the same coding performance and has practical applications.Secondly,Motion estimation is an important part of inter prediction,which takes almost 70%time of all inter prediction.Search strategy and matching criterion directly affect the efficiency and accuracy of finding the best matching block in the motion estimation.A fast search algorithm based on heterogeneous diamond pattern is proposed from the perspective of search strategy.Based on the analysis of the main factors affecting the performance of the TZSearch search algorithm and the distribution of the motion vector,a heterogeneous diamond is used to perform a rough search,and then vertical or horizontal hexagonal patterns are used to perform fine search for dynamically adapting to the image feature.The experimental results show that compared with HM 10.0,the algorithm proposed in this paper can reduce the total coding time by about 17%while maintaining the coding performance and speeding up the search process and shorten the time of motion estimation.