Algorithm Research And Optimization On High Efficiency Video Coding

High Efficiency Video Coding(HEVC)is the state-of-the-art international video coding standard from ITU/ISO/IEC.After finishing the development of HEVC standard,the application and optimization of HEVC become important research issues.The major contributions of this paper are as below:1.We analyzed the problems in the GOP level bit allocation of existing rate control algorithms,and deduced the general term formula of GOP level bit allocation of lambda domain rate control algorithm.To accommodate the new GOP structure of HEVC,we proposed two restrictions of GOP level bit allocation for HEVC constant bit-rate control.Based on these two restrictions,we proposed a new GOP level bit allocation for HEVC constant bit-rate control.The method matches the GOP coding structure of HEVC exactly,making the encoder buffer occupancy decrease linearly GOP by GOP and back to initial buffer state when finishing each Intra period.The bit allocation can improve rate control accuracy and R-D performance,and can also decrease buffer occupancy and bit fluctuation.In addition,we proposed a new measure for instant bit-rate error for HEVC random access configuration.2.Aiming at the GOP coding structure of random access configuration,we determined the importance of each GOP in a Intra period,and based on the importance we modified picture Lagrange multiplier calculated by rate control algorithm in a GOP.The method improves the R-D performance of rate control algorithm in random access configuration,and retains the rate control accuracy meanwhile.In addition,we explored the frame level bit allocation methods for random access configuration of new HM version where GOP size is 16,including fixed ratio bit allocation and adaptive ratio bit allocation.Exprimental results shows that the proposed two frame level bit allocation methods can both improve the rate control accuracy and rate-distortion performance greatly compared with the existing frame level equal bit allocation.3.In order to increase the coding performance when occurring scene change in low delay configuration,we analalyzed the features of pictures that occurring scene change,and then proposed a scene change detection algorithm.The scene change detection algorithm reach 100% detection accuracy in our test set;and based on the algorithm,we propose a strategy dealing with scene change in low delay configuration,which adjusts the GOP structure of sequence resulting in rate-distortion performance improvement no matter when the rate control is on or off.