A New Local Region-Based Active Contour Model for Image Segmentation Based on Adaptive Double Potential Well Function

In this paper, we present a modified local region-based active contour model employing an adaptive double potential well function for image segmentation. Initially, to circumvent the issue of the potential well function's excessive evolutionary pace within the zero potential well, which could lead to rapid level set evolution and inadvertent targeting of segmentation areas, we introduce an adaptive double potential well function. This function dynamically modulates coefficient by increasing the diffusion rate during the initial phase, decreasing it in the later stages, and mitigating it in the vicinity of the zero potential well. Subsequently, we incorporate a length term and a penalty term, both predicated on the adaptive double well function, into the energy functional derived from the local region-based Chan-Vese (LRCV) model. This integration serves to augment the edge smoothness of the curve and the precision of the segmentation process. Experimental outcomes demonstrate that our proposed model significantly augments segmentation accuracy when benchmarked against certain related models.