Video object segmentation based on HOS and multi-resolution watershed

Video object extraction is a key technology in content-based video coding such as MPEG-5. An image segmentation method for extraction of video objects from a video sequence is proposed by combining the temporal segmentation result with the spatial segmentation result. HOS (higher-order statistics) are used in the temporal segmentation, and a rough motion mask is obtained through post-processing. In the spatial segmentation, the multi-resolution watershed is proposed. Segmentation only based on motion information is unlikely to achieve an accurate result without the help of spatial information. So, the video object with accurate boundaries is extracted by combining the motion mask with regions segmented by using the multi-resolution watershed. The experimental results show that the proposed algorithm improves the performance of detection of moving objects.