Font Size: a A A

Research And Realization On Content-Based Video Retrieval

Posted on:2007-08-10Degree:MasterType:Thesis
Country:ChinaCandidate:S G LiuFull Text:PDF
GTID:2178360185493664Subject:Optics
Abstract/Summary:PDF Full Text Request
With the rapid development of optoelectronic, computer, internet and information technology, content-based multimedia browsing and retrieval has become one of the key researched fields of the information processing. As the most popular multimedia, video, consisting of image, audio and text, is an important processing object.Video retrieval technology, including digital image and video processing, has got a full improvement with the efforts of the researchers. However, it is also a challenge to develop an efficient and practical content-based video retrieval (CBVR) system.Based on the review of the related work done on the topic of CBVR, we improve two key techniques, shot boundary detection and structure indexing, and implement a CBVR system. The main content of this paper is as the following:1. A new shot boundary detection algorithm is proposed, based on machine learning on the compressed-domain. Motion vectors and residual energy extracted from source video show a well inter-frame characteristic change. Combining the temporal multi-scale frame structure and the prediction information, we proposed a new shot boundary detection algorithm. The key idea is that the features from compressed-domain are exploited by Support Vector Machine, and then shot boundaries are detected by the learned model. Experimental results show that the detection efficiency is highly improved while the precision is also good.2. The mechanism of the video structure indexing (VSI), which is to resolve the problem of the arbitrary accessing of the video data or the corresponding processed result, is further developed. It is based on the structure information of MPEG and video structuring such as shot boundary detection. And the VSI information is organized by using database. We implement real time CBVR under this VSI mechanism.3. We have designed and implemented a CBVR system, which is a complex project including many key techniques such as data parsing, video information mining and modeling, and the building of the mechanism of browsing and...
Keywords/Search Tags:Multimedia techniques, Shot boundary detection, Video indexing, Video structure analysis, Video content extraction, Video browsing and retrieval
PDF Full Text Request
Related items