Research On Intra Coding Algorithm Optimization For VVC

Nowadays,video has become an integral part of people's lives.Short videos,live broadcasts,and video conferencing continue to enrich people's daily lives.In order to improve the quality of video transmission,the fundamental approach is to improve the compression efficiency of the video and reduce the amount of data transmitted.In order to obtain compression capabilities beyond previous generations of video coding standards and have a high degree of versatility for videos,the next generation of international video coding standards,Versatile Video Coding(VVC),is being developed.It is expected to be released in 2020.At the current stage,VVC has paid the price of high implementation complexity though it has obtained excellent encoding performance.The intra prediction is an important module in the hybrid video coding framework.It makes full use of the spatial correlation of pixels to make predictions and improves compression efficiency.VVC uses more sophisticated intra prediction technology,which not only improves traditional intra coding tools such as angular intra prediction,but also adds many new intra coding tools.In the process of participating in the formulation of VVC,this paper focuses on the intra coding tools and proposed optimization algorithms.Three of them have been adopted by VVC.The intra coding tools studied in this paper include DC mode,CrossComponent Linear Model(CCLM)and Matrix based Intra Prediction(MIP).The specific research contents and innovation works are as follows:1.A position-based reference pixel downsampling algorithm is proposed.The algorithm can reduce the number of reference pixels and reduce the computational complexity of some intra coding tools.The algorithm is applied to three intra coding tools to verify the effectiveness.(1)A DC optimization algorithm based on pixel position is proposed.The proposed positionbased reference pixel downsampling algorithm is applied to the calculation process of DC mode after analyzing the implementation process of DC mode in VVC.The algorithm reduces the computational complexity without affecting the coding performance.(2)A CCLM optimization algorithm based on pixel position is proposed.The proposed position-based reference pixel downsampling algorithm is applied to the linear model parameter derivation process of CCLM after analyzing CCLM coding tool in VVC,which reduces the computational complexity and improves the coding performance.An application of this algorithm has been adopted by VVC and has been integrated into VVC reference software.(3)A Two Step Chroma Prediction Mode(TSCPM)optimization algorithm based on pixel position is proposed.The TSCPM coding tool in the Audio Video coding Standard 3(AVS3)in China has been explored at first.Then the proposed position-based reference pixel downsampling algorithm is applied to TSCPM which improves encoding performance without increasing the computational complexity.2.The MIP coding tool in VVC is analyzed,and two optimization algorithms are proposed.(1)A classification-based MIP optimization algorithm is proposed.Through the modification of the MIP classification,the implementation process of MIP is simplified and the coding performance is hardly affected.This algorithm has been adopted by VVC and has been integrated in VVC reference software.(2)A mapping-based MIP optimization algorithm is proposed.Through the fixed mapping of MIP,the fusion process of MIP and other intra-coding tools is optimized and the coding performance is hardly affected.This algorithm has been adopted by VVC and has been integrated in VVC reference software.