| Summary: | 碩士 === 中華大學 === 資訊工程學系碩士班 === 91 === Recently, more and more complicated digit video communication services and devices emerged, and these services such as online game, video on demand, and interactive TV are available. Those techniques adopt the design method based on object-oriented approach. Furthermore, the new video coding standard MPEG-4 is involving content-based functionalities that decompose a video sequence into one or more video object planes (VOPs), and each VOP means one object. In the traditional segmentation method, we need high computation if we want to get better-segmented result, so a good segmentation algorithm is important to generate object in a video sequence.
In this thesis, we propose an automatic segmentation algorithm that can extract the moving object efficiently, and the algorithm is based on combination of spatial and temporal information. For spatial information, we use color edge extraction method to extract edge. For temporal information, we use change detection method to locate moving region. Furthermore, we use contour information instead of region information, and to improve the performance of segmentation algorithm. In the experiments, we analyzed the PSNR and compression ratio by MPEG-4 reference software for the proposed segmentation algorithm, and the object segmented by the proposed algorithm achieve better quality and high compression ratio.
|
|---|
