Research On Fast Intra Prediction Algorithm In H.264 Based On Template And Threshold

With the popularization of broadband networks and the development of multimedia technology,the digital video is getting more and more important in our daily life.But,due to the limitation of the network,the business of digital video encounters a bottleneck.In 2003, Joint Video Team(JVT) came out with a new standard,advanced video coding part of the final draft,which was called H.264/AVC standard by ITU-T or advanced video coding part of the final draft of MPEG-4 by ISO/IEC.The new standard becomes a standard offering a higher compression ratio and friendly network interface.The standard contains a lot of advanced video compression coding technologies and thoughts,which is noticeably more improved than the previous video coding standard in the coding efficiency.However,the introduction of these new methods of video coding highly increased the computational complexity of H.264/AVC,ten times more than the simple configuration of MPEG-4.This greatly restricts the dissemination and use of H.264/AVC.So we should optimize the standards to get higher efficiency,especially in Real-Time application.This thesis offers an introduction to the basic principles and key technologies of H.264/AVC.It focuses on intra-prediction encoding and analyses the computational complexity of the intra-prediction and at the same time analyses the mode decision in 4×4 macroblock and 16×16 macroblock and the correlation of the adjacent block.According to the information of the adjacent block,this paper makes a new arithmetic which can cut down the unnecessary modes in intra-prediction and improve the coding efficiency.The main work in this thesis covers the following aspects:1.Based on the analysis of the research of intra-frame coding,the paper puts forward a new method which uses the template technique to decide the prediction mode.In 4×4 block we use the eights direction template to decide the direction.In 16×16 block we use the scalable template to decide the best prediction mode.With the two methods this arithmetic can decrease the prediction modes and reduce the computational complexity.2.Using the fast arithmetic to judge prediction mode of the current macroblock.For those intense macroblocks or plane macroblocks,it only uses the prediction of 4×4 block or 16×16 block to raise the coding efficiency.Compared with the full research arithmetic,it decreases the unnecessary style of macroblocks and reduces the computational complexity.3.This thesis analyses the correlation of the adjacent block and puts forward a new arithmetic based on the correlation of the adjacent block to predict the current block prediction-mode.Using the reference software JM13.2,this paper compares the new arithmetic with the full search arithmetic.Experimental results show that the proposed methods can reduce encoding time with less performance degradation and bit-rate increase.