Study Of Efficient High Dynamic Range Video Coding Algorithms

With the development of video technology,the requirements of High Dynamic Range(HDR)video bring great challenges to the collection,compression and transmission of videos.The legacy codec is not aimed at HDR videos,and the compression efficiency of HDR video coding is low.As an emerging video content,most displays cannot support HDR videos,and the transmission system needs to provide both HDR videos and Standard Dynamic Range(SDR)videos.The scalable video coding framework can encode both HDR videos and the corresponding SDR videos,but the complexity of encoding is doubled.HDR videos utilize a new data format with a bit depth which is more than 8 bits.In most HDR videos,the distribution of chromaticity component is so concentrated that many available codewords are not utilized.By recording the chromaticity component with more available codewords,the chromaticity information of the original video can be more finely preserved.We studied the compression of HDR videos,and the main research results are as follows:1.A fast coding algorithm for HDR videos within a scalable framework is proposed.Firstly,this paper studies the generation process of SDR videos and HDR videos and characteristics of the generated video.Then,this paper analyzes the correlation between the optimal Coding Unit(CU)results of HDR videos and the corresponding SDR videos in coding.Finally,this paper proposes an algorithm to determine the CU depth range using the adjacent CU information of HDR video and the co-located CU information of the SDR video.At the same time,this paper also proposed a Coding Tree Unit(CTU)depth range prediction algorithm based on support vector machine.Support Vector Machine(SVM)is commonly used for classification,so each CTU is divided into three categories according to the depth distribution of its internal CU in this paper.The training set of SVM consists of features that affect the results of CU depth and the CTU type to which each CTU belongs.Experiments show that the proposed algorithm greatly reduces the omplexity of HDR video coding within the tolerable loss of BDBR performance under intra coding.2.Based on HDR10 framework,an adaptive codeword allocation algorithm based on chromaticity distribution is proposed.Firstly,this paper introduces the codeword allocation of video coding framework HDR10.This paper introduces the non-uniform distribution of pixel values in YCb Cr color space and proposes an adaptive codeword allocation algorithm.This algorithm divides the whole chromaticity range into 64 small ranges uniformly and calculates the cumulative quantization error of pixels in each small range and the cumulative quantization error in the whole range.At the same time,ratios of each small range cumulative error on the whole range cumulative error are used for codeword allocation.After analyzing the influence of upper and lower limit of codewords,this paper determines the upper and lower limit through a lot of experiments.In the test,the algorithm improves the utilization efficiency of codeword and greatly improves the coding efficiency under intra coding.