Research On Low Complexity Algorithm In High Effciency Video Coding

With the development of communication technology and multimedia technology,video applications have been pervasive in our lives.The informations get from videos are very directly.But it can not be transmitted and stored before compressed for its huge amount of data.New standards have been established to meet the further requirement of video coding.In 2013,the next generation video coding standard named High Efficiency Video Coding(HEVC)has been published by the JCT-VC.The main goal of HEVC is to deliver about 50% higher compression performance relative to H.264/AVC with equal perceptual quality.HEVC adopts many new methods to improve coding efficiency.However,these techniques also bring great burden on video encoder.So research on reducing the video coding complexity is very crucial.This disseration focuses on the low complexity algorithm in HEVC.The improvements include inter prediction,intra prediction,sample adaptive offset and entropy coding.The main innovations are listed as follows:(1)An effective coding unit splitting algorithm in inter prediction of HEVC is proposed.Based on the motion homogeneity and the quad-tree structures of coding unit,a certain threshold is used as the coding unit split flag.By skipping some specific sub-CUs with less motion informations,the coding complexity will be dramatically decreased.Experimental results show that the proposed technique can save 46.1% coding time in average with negligible loss of coding efficiency.(2)By analysing the percentage of coding time with each CU depth and the relationship between CU depth and visual saliency of videos,a complexity control algorithm of HEVC with visual saliency is proposed.Based on their visual saliency characteristics and the target level,the proposed algorithm adjusts the maximum coding depth of each coding unit.The saliency threshold is used to decide whether the coding unit should be split into smaller CUs.By this way,the coding complexity control can be achieved with minimal loss of visual distortion.(3)This dissertation proposes two algorithms of sample adaptive offset to reduce the coding complexity.According to the relevance of sample adaptive offset values and CU depths,a pre-decision SAO method based on CU depth is proposed.It can skip some CUs by classify the video textural features.The key idea of the second algorithm is to decrease the number of SAO candidates by certain visual saliency threshold.Both of them can effectively reduce the coding time of SAO processing.(4)Based on the characteristic of context-based adaptive binary arithmetic coding and bypass coding,an entropy algorithm of transform coefficients coding is proposed.And statistics show there are many zeroes and small absolute value transform coefficients.According to these features of transform coefficients and entropy coding,an improved entropy coding algorithm for transform coefficients is proposed by reducing some context adaptive binary arithmetic coding bins.Simulation results are provided to quantify the complexity reduction.