Research On Video Feature Acquisition And Efficient Coding Method

In recent years,with the rapid development of network technology and streaming media technology,high definition(HD)video and even ultra-HD video have gradually entered people's lives.This trend leads to huge demand for video data processing and video transmission.At the same time,the research on HD and ultra-HD video compression technology grows fast,and now High Efficiency Video Coding(HEVC)has been improved further.Compared with the previous generation H.264/AVC standard,HEVC greatly improves video compression efficiency without degrading the quality of video reconstruction.In HEVC standard,it follows the hybrid coding framework proposed in the H.264/AVC standard,and adds a transform-quantization module,an improved intra and inter coding technique,and a loop filtering technique to efficiently compress HD video.However,because of the complicated recursive process and its calculation,the overall complexity of HEVC coding greatly increases,and the compression time is raised by nearly 50%compared with H.264/AVC.It can not meet the needs of real-time video transmission.Therefore,this paper proposes two optimization methods for the problem:1.This paper proposes a fast CU partition decision algorithm based on feature extraction and Support Vector Machine(SVM)in order to reduce the high computational complexity of intra-frame coding in the HEVC standard.It mainly contains two major steps.Firstly,acquiring the CU features in the original coding process,and using the CU features is to establish prediction models using the machine learning SVM method.Then the model is applied to the optimization algorithm,the same type of features are extracted for the current coding block,then input into the SVM prediction model to obtain the prediction result for fast coding.In order to further reduce the computational complexity,the depth range constraint and rate distortion cost prediction method are added to the optimization method,which improves the prediction accuracy of the algorithm and reduces the compression time further.The experimental results show that this method can achieve the desired effect,it can reduce the compression time by 42.11%and the bit rate only increases by 1.93%.2.This paper proposes a fast CU partition and mode decision based on feature extraction and decision tree for Screen Content Coding(SCC).SCC is actually an extended coding standard of HEVC.It is mainly used for computer-synthesized video,such as mobile devices and virtual desktops.Since SCC adds some new modules in the screen video based on HEVC,the computational complexity is greatly increased.The method mainly includes three parts.Firstly,the feature information of the coding block is extracted in the original coding,and used to train and verify the decision tree model.Then,the prediction accuracy of each node of the decision tree model is calculated.The decision tree with node accuracy as well as the prediction information of neighboring blocks is used to predict the SCC coding unit partition process and the predict unit mode selection process,so as to skip the unnecessary CU depth decision and mode decision.Experimental simulations show that the fast algorithm can achieve 33.9%time savings,while the bit rate consumption is negligible.