A global minimization hybrid active contour model with applications to oil spill images

Active contour model is popularly and widely used in the field of image segmentation, which is based on superior theoretical properties and efficient numerical methods. Nevertheless, one of the prominent disadvantages of this kind of model is the existence of local minima in its functional energy. In this paper, we propose a novel global minimization hybrid active contour model. This model effectively integrates the edge information, the local region information and the global region information of the image, which is relatively sufficient to extract the object boundaries. Furthermore, we introduce an efficient and fast numerical approach to globally minimize the proposed model, which is through a dual formulation of the minimization problem and easy to implement. The proposed model is robust enough to the initial condition and does not need to initialize the contour in a distance function and re-initialize it periodically during the evolution process. Specially, we applied the proposed model to segment oil spill images, in which there usually exist the noise, blurry boundaries, and intensity inhomogeneity. Compared with the state-of-the-art models, experiment results demonstrate the performance and effectiveness of the proposed model with applications to synthetical and real images, especially for oil spill images.