Abstract: The paper proposes a method of images contour extraction which combines the statistical method and active contour models(ACM).To choose the contour points of points distribution models automatically, the minimum description length was chosen as the object function.A typical contour model is obtained from an image training set.After the initialization of a new image's ACM contour, the typical contour mentioned above is used as the deformation constraint to extract the nearly optimum contour.The experiments with the real images show the effectiveness of this method in comparison with the traditional active contour models method.

Key words: contour extraction for medical images, active contour models, Snakes models, points distribution models, minimum description length