A novel foreground/background decision using in unsupervised segmentation of moving objects in video sequences

This paper presents an unsupervised segmentation technique for separating moving objects from background in image sequences. This technique utilizes the spatial and temporal information. For localization of moving objects in the video sequence, two consecutive image frames in the temporal direction are examined and a significance test is performed from two consecutive difference images. Spatial segmentation is performed to divide each image into semantic regions and find precise object boundaries of moving objects. In order to achieve high performance in segmentation, a novel foreground/background decision technique is proposed. Our technique includes two methods for different situations. Method one is applied based on two distinct thresholds for foreground/background decision. Method two is used by a sorting of changed-pixels ratio to get a more precise decision. The results show a good segmentation performance becomes achievable in generic images, and enrich the MPEG-4 and MPEG-7 for content-based indexing and retrieval of multimedia applications.