Active contour model, also shown by the nickname, snake, is a powerful tool for image segmentation. Since it provides continuous boundaries for regions of interests, a snake model has the advantage over edge-detecting approaches for segmentation in that no edge linking would be required. Powerful as it is, the snake model suffers several cons, e.g., corner- fitting problems, easily trapped by noise pixels, etc., which limit its usage, especially, on a noisy image. With the ultimate goal to accomplish segmentation on a noisy image composed of arbitrary corner shapes, this paper proposes a new snake model, namely, evolutionary snake-balloon model, to solve the corner-fitting problem, which is a common problem in most snake algorithms. The key idea is to decompose the snaxels (control points) into subsets and perform the snake deformation one subset after another. The evolutionary model is an effective approach to catching corner points while the snake is moving toward the desired boundaries. In addition to the evolutionary concept, a new adaptive methodology is proposed in this paper to determine the weighting factor of each energy term such that the snake may cope with local minimums more effectively. Experiments have been carried out on synthetic images with random noises and ultrasound images to verify the proposed evolutionary snake-balloon model.
|頁（從 - 到）
|Biomedical Engineering - Applications, Basis and Communications
|Published - 25 4月 1998