A Two-Phase Snake Method for Object Segmentation against Background Clutter and Boundary Concavities

Abstract

Various object segmentation methods have been proposed based on the classical Active Contour Model (Snake), which has been used extensively to locate object boundaries in images. However, these methods have limited capability in overcoming the problems of background clutter and boundary concavities. Background clutter has a noise, which obstruct the snake moving and determining the Object Of Interest (OOI), also the snake ability suffers to move into boundary concavities. In this research, we propose a new snake method that can perform more efficiently in the presence of background clutter and boundary concavities. Our approach will use two- phase snake instead of a greedy snake algorithm, which works by computed energy functions on neighborhood around each snake point , then move to the position with the lowest energy, and it will use scale space continuation to increase the snake ability to find the OOI contour from cluttered background. The first snake-phase try to converge on edges until stopping but that doesn't mean founding the OOI boundary. After that the second snake-phase starts its completing until the boundary of the object of interest is found. The two- phase snake method is testing on a number of different images under most conditions. The experimental results errors also will be compared on the two- phase snake. Performance of the new method will be evaluated in terms of accuracy and performance compared with existing methods, which are based on the classical model.