Motion detection based on contour strings

A novel method for motion detection is presented which combines three classical approaches. A modified image subtraction approach suppressing local lighting changes is used to determine contour point strings of moving objects. Thus the pixels processed by the proposed algorithms are restricted to contour points which increases computational efficiency. Initially we process the scene without moving objects to determine all contours appearing in the background image. Then the contour strings from the input image are compared with the background contours to separate the moving subcontours from the static ones. The proposed algorithm is stable for illumination changes and can be computed in real-time.