A method for multiple rigid-object motion segmentation based on detection and consistent matching of relevant points in image sequences.