Font Size: a A A

For Mpeg-4 Video Object Segmentation Algorithm

Posted on:2004-08-24Degree:MasterType:Thesis
Country:ChinaCandidate:L LuoFull Text:PDF
GTID:2208360095460188Subject:Signal and Information Processing
Abstract/Summary:PDF Full Text Request
With advances in communication and information processing technologies, the video-driven application show a very large degree of flexibility and extensibility. Visual communication is the fastest growing vehicle for information. A lot of digital applications and services are emerging, such as: digital TV, teleconference, videophone and interactive multimedia. These diversified applications and services with a large amount of data demand more advanced digital signal processing technologies for efficient memory and transmission, accurate analysis and flexible manipulation.MPEG-4 standard is such a standard to solve above problems. Audio-video object-based coding and description are the distinct features of MPEG-4 standard. In the MPEG-4 standard, video frame was composed of a series of independent semantic video object, which was encoded independently. So, the performance of the segmentation algorithm is crucial to the final MPEG-4 coding products. Video segmentation is a challenging topic of research, no single algorithm has been reported in the literature that is generally applicable.In the dissertation we proposed a spatial and temporal VOPs' Segmentation and Extraction methods for MPEG-4. I concentrate my research work on accurate , common and automatic algorithms, At the same time reduce the complexity and computation burden of algorithm as could as possible to match the needs of real-time requirement. This paper was organized as follows: First, an overview about the multimedia technology, MPEG standards and some research backgrounds were shown. Second, some popular spatial and temporal technology about the VOP segmentation were introduced. Third, a modified watershed segmentation preprocessed by multi-scale morphological filters and post processed with improved fast region merging methods was proposed to obtain a good spatial segmentation . Forth, to extract the semantic object from the spatial segmentation, we applied a temporal segmentation based on the global motion compensation, statistical model-based change detection and post processing technology to get the motion mask. Fifth, a methods based on the above spatial and temporal result was brought forward to obtain the final, a two threshold method was proposed. Finally, we summary our algorithm and propose the improve direction in the future .At each stage of our algorithms, some experimental result was given to validate the performance of the proposed algorithms.
Keywords/Search Tags:MPEG-4, Video object segmentation, Morphological filters, Watershed segmentation, Change Detection
PDF Full Text Request
Related items