Font Size: a A A

Study On Video Object Segmentation And Its Realization Method On CNNUM

Posted on:2006-02-22Degree:DoctorType:Dissertation
Country:ChinaCandidate:Q L ZhangFull Text:PDF
GTID:1118360155460337Subject:Communication and Information System
Abstract/Summary:PDF Full Text Request
Classical video coding standards such as H.26x and MPEG-1/2 are frame-based techniques, and no segmentation of video scenarios is required. Their high compression performance makes them widely used in video application. With the proliferation of multimedia information, people are no more satisfied with simple navigation of video contents, but require object-based functionalities.Therefore, Motion Picture Experts Group (MPEG) published the secondary generation video coding standard, MPEG-4. Compared with the first generation standard, a significant character of MPEG-4 is object-based coding which means to code videos as a set of semantic video objects. Video object segmentation is necessary to get each video object. Video object segmentation has been promoted greatly by the occurrence of MPEG-4 but is not limited to MPEG-4. Video object segmentation can serve for many application in computer vision area. Some typical applications of video object segmentation are video coding, video authoring and edit, video retrival, video-based monitoring, etc.Video object segmentation is a nut in video processing and computer vision. The diffculty of video object segmentation lies on two aspects. One is the extreme complexity of video scenes, which means no uniform model for all video objects. Another one is the definition and description of semantic video object. Video object segmentation is carried out on low vision level while semantic video objects are defined on high vision level. It is difficult to get high-level objects by low-level segmentation. Generaly, there are two problems in the current segmentation methods: one is no universal algorithm suited for all the scenes, another one is most of the current algorithms are hard to meet the real-time performance.This dissertation focuses on the methodology and techniques for video object segmentation under the framework of MPEG-4, cellular neural network is introduced to conquer the problem of real-time performance.Major work of this dissertation is as follows:First, a face extraction algorithm based on edge projection is proposed. An important character is discovered via the analysis of head-shoulder sequences, the motion details are rich in the face region, so the approximate coordinates of the face can determined by the vertical and horizontal projection of DFD image. Then this...
Keywords/Search Tags:video object segmentation, MPEG-4, spatio-temporal, mathematic morphology, cellular neural networks, optical flow, watershed
PDF Full Text Request
Related items