A Pyramidal Approach to Active Contours Implementation for 2D Gray Scale Image Segmentation

Active contours or snakes have been widely used for segmenting objects of interest from image background. Among all the types of developed parametric snakes, GVF snake and its variations have been proved effective in terms of large capture range, convergence to concavities and immunity to noise etc. Even though being effective, when such a snake is applied on a high resolution image, it takes considerably large number of iterations to converge to the boundary and might be poorly converged to the concavities due to bad selection of initial contour. To overcome such issues, in our proposed method, we have used a pyramidal multi-resolution approach and implemented the snake on the lowest resolution image and subsequently on the highest level images in the pyramid. The method is formulated, implemented and tested over a number of 2D gray scale images. Experimental results show that our method is able to reduce the number of iterations effectively while giving a better segmentation.