Boundary detection can be considered as a kind of optimization in convergence toward the optimal estimate boundary. Active Contour Model, "Snake", is one of the most popular boundary detection methods, and has a principle of an energy-minimizing spline to estimate the closest contour of a target object in an image gradually from an initial contour. Therefore, Snake is a good example for improvement of boundary detection by general optimization techniques. We have proposed Multi-Snake, integration of Snake with decentralized cooperative processing, and verified its effectiveness. This paper focuses on cooperation of Snake and a simple model-based Snake, which is based on the idea that the initial contour contains some information on the target shape, and this information can be used as a model in the process of boundary detection. This method addresses Snake's drawback detecting the very complex boundary, in particular concave boundary. We verify effectiveness of our proposal through some experiments using artificial images and real images.
展开▼
