A fast algorithm for active contours

A method of controlling snakes that combines speed, flexibility, and simplicity is presented. It is compared to the original variational calculus method of M. Kass et al. (1987) and the dynamic programming method developed by A.A. Amini et al. (1988) and found to be comparable in final results, while being faster than dynamic programming and more stable and flexible for including hard constraints than the variational calculus approach. The introduction of the concept of curvature highlights the problem of how to approximate curvature when a curve is represented by a set of discrete points. The advantages and disadvantages of a number of different approximations of curvature are pointed out.<<ETX>>