Moving object segmentation for video surveillance and conferencing applications

Video surveillance and conferencing systems have an impressive spread both for their practical application and interest as research issue. The common approach used in such, systems consists of a good segmentation of moving objects. This paper presents an algorithm for segmenting and extracting moving objects suitable for surveillance and videoconferencing applications, where a still background frame can be captured beforehand. Since edge detection are often used to extract accurate boundaries of the objects in the scene, the first step in our algorithm is accomplished by combining two kinds of edge points which are detected from the frame difference and the background subtraction. After removing edge points that belong to the background frame, the resulting moving edge map is fed to the object extracting step. A fundamental task in this step is to declare the candidates of the moving object, followed by applying morphological closing and opening operations. The algorithm is implemented on a real video sequences and good segmentation performance is achieved.