Research On Transform Coding Technology For High Efficiency Video Coding

Since the 20th century,with the development of computer technology,network technology and multimedia technology,the telemedicine,distance education,video conference and other applications based on video technology are also developing rapidly.Video data is going to be large data volume and high-resolution.The transmission and storage of heavy video data challenge the development of video technology.The devel-opment of efficient video compression technology has became the bottleneck of video technology development.In May 2003,JVT,a joint coding group composed of the Motion Picture Experts Group(MPEG)and the Video Coding Experts Group(VCEG),released the current mainstream video coding standard named H.264/AVC.Compared to the previous coding standards,the compression performance of H.264/AVC is in-creased by 50%.After about 10 years,in January 2013,JCT-VC,a joint coding group composed of MPEG and VCEG,released a new generation of video coding standard HEVC/H.265 which outperforms H.264 about 50%in coding efficiency.The existing video coding standard removes the temporal and spatial redundancy by prediction,the prediction error is represented as the residual,the residual signal is then subjected to entropy coding after transform and quantization.While the amount of residual data occupies the vast majority of the total amount of encoded data,a good transform will greatly improve the compression efficiency of the encoder.Whether it is H.264 or HEVC,the transform uses fixed transform kernel,but different image texture result in different statistical characteristics of the residual.It is clear that the correlation can not be effectively eliminated by performing the same transform kernel for the residuals of different statistical characteristics.This article does two aspects of the study on transform coding.On the one hand,the efficiency of transform coding is improved by increasing the number of transform kernels.On the other hand,the efficiency of transform coding is improved by finding an adaptive transform kernel based on the content of the transform unit.Specifically,the main innovations of this paper are as follows:In this paper,the adaptive multi-transform technique is proposed.The transform kernels DCT-V,DCT-VIII,DST-I are added to the existing coding framework.The encoder will find a transform kernel that is most suitable for encoding by trying differ-ent transform kernels.The adaptive multi-transform technique provides more adaptive transform choice compared to the existing coding framework.This paper also proposes a fast adaptive multi-transform technique based on the analyzing of the statistical data.Experiments show that the adaptive multi-transform algorithm proposed in this paper can improve the compression efficiency by about 2%-3%.In this paper,orthogonal matching pursuit technique is also proposed.Compared to the transform kernel of existing coding framework,the adaptive transform kernel is obtained by template matching and online training.We will find an optimal transform method for transform coding by comparing it to the existing coding framework.The orthogonal matching pursuit transform takes the diversity of the video content into ac-count,as compared to the existing coding framework which uses the same transform kernels for all video content.Experiments show that the orthogonal matching pursuit technique proposed in this paper can improve the compression efficiency by about 1%?3%.