An active contour model algorithm combined with anisotropic diffusion filtering and global pre-fitting energy

Abstract

Traditional local active contour models mainly depend on the data fitting term to obtain image edge information, which are sensitive to the position of initial contour. In fact, setting an unreasonable initial contour will result in a time-consuming image segmentation process and difficult evolution at weak boundary. To solve this problem, we come up with a second-order edge detection operator which is based on the global gradient image information to replace the traditional data fitting term, and to obtain edge information with higher speed and accuracy. In addition, the anisotropic edge enhancement method has been introduced to improve the robustness to the weak boundary. In this way, an active contour model which is insensitive to initial contour position is proposed to achieve a faster and more accurate image segmentation result.