| Summary: | 碩士 === 國立中興大學 === 電機工程學系所 === 94 === The segmentation of video objects is an important research topic in digital video processing. Due to the unreliability of object motion information and the lack of higher level guidance, video objects segmentation is still a challenging topic. Since the approach suitable for general situations is almost not feasible, most of the researches focus on the video object segmentation with constraints depending on different applications to obtain reasonable results. In the same view, we restrict video objects to foreground objects and background objects. Therefore, the purpose of this thesis is to extract foreground objects with motion and the static background objects. The method we used includes two parts: the initial segmentation and the MRF region classification. First, the watershed algorithm is used to segment the image to acquire regions with spatial coherence. Second, the Markov Random Field (MRF) approach is adopted to classify the regions into the foreground and the background regions. This approach is designed to process the QCIF or CIF video sequences in the YUV format.
|
|---|
