Abstract: A method for segmenting an image in which an arbitrarily shaped contour is deformed stochastically until it approximates the contour of a target object. The evolution of the contour is controlled by a simulated annealing process which causes the contour to settle into a global minimum of an image-derived "energy" function. The non-parametric energy function is derived from the statistical properties of previously-segmented training images. High computational complexity is avoided by using an efficient method of introducing a random local perturbation, and assuring the resulting shape changes are unbiased. This method for perturbing the contour allows for execution times several orders of magnitude shorter than in simple implementations.