GAC-based color image selective segmentation under geometrical constraints

Color image segmentation is a very important and practical problem of image processing. Though conventional variational segmentation models are a kind of high efficiency and widely used segmentation models, however, these models segment image globally, as for tasks to segment one object of a given image, these models will fail. In order to overcome above shortcoming, this paper presents a new model which can segment a color image selectively. It is a GAC (Geodesic Active Contour) based model for color image with geometrical constraints. This model consists of color image edge detection function, distance function and color image Geodesic Active Contour (GAC) model. In order to speed up the segmentation, this paper applies the AOS (Additive Operator Splitting) method. Experiments on synthetic and real images illustrate the validity of the proposed model.