A study of medical image segmentation technique using active contour model based on morphological gradient: with some synthetic images

Abstract

Snake, also known as the Active Contour Model, is an actively developing research area for the image segmentation algorithm. Gradient Vector Flow (GVF) Snake resolved the problems associated with the initialization and concave region resulting from using the gradient vector flow as the external force. However, the problem resides in the use of Gaussian filtering that result in blurring effect on the object boundary. Consequently we have difficulties to find exact contour of the object boundary. In order to resolve this problem, the morphological gradient was used for a new edge map to create an external force more precise than that formed through the GVF Snake. For this experiment, we used three different types of synthetically generated images. All of the comparison tests were carried out under the same conditions (i.e. with same parameters in the GVF Snake algorithm) that result in the GVF Snake made the optimal movement. Even though, the improvements of our algorithm are clearly observed in the results. We evaluated the results with estimation error and minimum distance error.