Font Size: a A A

Research On MPEG-4 Video Algorithm Optimization

Posted on:2007-09-08Degree:MasterType:Thesis
Country:ChinaCandidate:G Y HuangFull Text:PDF
GTID:2178360245484001Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
MPEG-4 video coding standard is the first standard which based on the second generation technology; it is a new standard which the content of video can be interacted. It supplies shape coding, motion estimation and compensation, texture coding, sprite coding, scalable coding for video object plane. MPEG-4 focuses on the interaction and operation of the content of image. VOP and sprite are important concept in MPEG-4, and they are important methods which implement low bit rate transfer. This paper study dynamic Sprite generation and MPEG-4 coding algorithm furtherly.This paper study furtherly the two key steps in sprite generation. They are global motion estimation and sprite image mosaicing, and describe the mothods in details. Global motion is modeled under 8-parameters perspective model. When sloving the global motion parameters, three-layer pyramid images are generated firstly using 3-order filter. Following, a typical tree-step block-maching searching algorithm is used to estimate tranformation components on the top-layer image. And then compute the perspective model parameters using iterative optimization gradient descent algorithm in every layer pyramid image. Lastly, get the global motion parameters when finishing iterative computation on the base image. This paper describes the global motion estimation under perspective model and sprite generation in details. Considering the effect of noise on global motion, this paper proposal an improved algorithm which based on robust M-estimator. In order to generate sprite real time, the improved algorithm employs temporary-domain and spatial-domain feature point selection module to select the feature points which contribute more to global motion to anticipate in iterative compution.Next work of this paper is to generate an excellent visual quanlity sprite. To avoid the pixels which belonges to foreground video object miss register in sprite and the ruin the sprite. Considering the difference reliability of the pixels which need to register in the sprite, this paper register the pixels based on their reliability to generate the sprite image. Only the pixels belonged to back ground video object need to register, diffderent reliability applied different mosaicing strategy.
Keywords/Search Tags:global motion estimation, M-Estimator, sprite generation, image mosaicing, block match estimation
PDF Full Text Request
Related items