The Motion Estimation Module Modeling And Simulation Of MPEG-4Video Processing System

Motion estimation algorithm is one of the most critical technologies in theMPEG-4video compression framework. It is of significant impact on accuracy andefficiency of video coding. Its main purpose is to increase the compression ratio largelyby using the correlative motion information between adjacent frames and using thereference frame and the residuals frame to replace the current information. Motionestimation is large amount of calculation, accounting for more than80%of the totalvideo encoding time. A fast algorithm for motion estimation algorithm is particularlyimportant.This paper mainly discusses the motion estimation part of MPEG-4. The matchingcriteria, search strategy and the determination of initial search points were analyzed andcorresponding improved. We have further study of search algorithm which has theimpact on the efficiency of the motion estimation. Analysis and comparison of severalcommon search strategies: three-step search, diamond search, hexagonal search, etc; acareful study on video features. Then presents a hybrid search strategy of hexagonal andasymmetric cross with search trends, through search for defects in hexagonal anddiamond search algorithm: no use of the correlation of adjacent macro blocks. Bycomparing PSNR and executive time, to verify the correctness and efficiency of thealgorithm, so that the compression ratio of the system, the signal to noise ratio toachieve design requirements.Based on the study of heterogeneous systems-level modeling and simulationplatform PtolemyII, design motion estimation actor and reverse actor which based onhybrid search strategies of hexagonal and asymmetric cross combination with searchtrends algorithm. Respectively create model for motion estimation actor and reverseactor, and model and simulate in the SDF domain on PtolemyII platform.