Tracking multiple objects using moving snakes

Abstract

Active contours or snakes are widely used for segmentation and tracking. The ability of a snake to track an object depends on the movement of the object. If the object moves too far from one frame to another, the snake risks losing the true contour location. The subsequent evolution steps are negatively affected, reporting a false contour that can propagate to other frames. To overcome this problem a new snake algorithm has been developed. This new technique, moving snakes, works in two steps. During the fist step, the snake is translated as a rigid body towards the contour. This translation is calculated using the external force field of the image, therefore it does not require prior knowledge about the object movement. In the second step the actual shape evolution of the snake takes place.