Object contour extraction from motion parameters

An approach to contour extraction using motion parameters derived from optical flow techniques is presented. The velocity vector, which represents 2D surface motion of a sequence of images, is used to form a velocity-direction distribution. The components of the velocity vectors determine its location in the distribution space. Velocity vectors that are orientated in the same direction and with a similar velocity form an object cluster. Each object cluster is used to find the velocity and direction of the moving object. The region containing the object clusters is used to define the object's contour. Contour extraction using motion parameters is fast and accurate in analysing complex sequences that inherently contain noise. The technique is shown to extract separately the contours of two differently orientated objects from a test sequence. The extraction of an object's contour in a complex sequence is also presented.<<ETX>>