A novel Accelerated Greedy Snake Algorithm for active contours

In this paper, we propose a new Accelerated Greedy Snake Algorithm (AGSA) for faster convergence of the active contour optimization problem. The new algorithm takes advantage of the similarity in image pixel gradients to take larger steps in the initial stages of the snake. Due to its fast convergence, the snake can be initialized far away from the object without any issues. This algorithm also uses some intelligent techniques (e.g. re-sampling, relaxation) to maintain a regular shape of the snake at all times while approaching the final contour. Experimental results on three test cases are presented, where the convergence efficiency of our method has been compared with three contemporary algorithms in terms of number of iterations and computational time.