Improved BVF Snake Model for the Segmentation of an Image with Multiple Targets

The parametric active contour model (Snake) has a wide range of applications in the field of computer vision and image processing. The traditional Snake model has many shortcomings, including high computational effort and failure to converge to the specific target when multiple targets exist. Although the boundary vector field (BVF) Snake reduces the computational intensity, it still fails to converge to the specific target when there are multiple targets. In this paper, the force field is in-spected to determine the reasons why the BVF Snake is not applicable to the segmentation of an image with multiple targets, and improvements are made to propose an improved directional BVF (DBVF) Snake model. In this model, a directional vector field is obtained by adding one point inside a specific target to determine the convergence direction of the Snake, so that it can converge to a specific target in the presence of multiple targets. In the paper, the DBVF Snake model is compared with the common Snake models in two aspects of increasing the capture range and adding different noises, and the sonar images are segmented and segmented results show that the DBVF Snake model is valid for the segmentation of an image with multiple targets.