Active Contour Model Based on Intensity-Spatial Information Fusion for Inhomogeneous Image Segmentation
Meng Zhang, Yi Yang, Sixian Zhang, Pengbo Mi · 2025
Active contour model is a common image segmentation method, but it usually performs poorly in the inhomogeneous image segmentation. To address this problem, a novel active contour model based on intensity-spatial information fusion is proposed. First, it can dynamically adjust the weight between the intensity information in the global and local regions according to the inhomogeneity degree around the evolution curve to simultaneously enhance the robustness to the initial evolution curve position and intensity inhomogeneity. Then, this paper uses the superpixel segmentation and kernel density estimation to obtain the fuzzy edge response and embeds it into the spatial prior information to eliminate segmentation noise and avoid oversmoothing. Finally, this paper verifies that the proposed model can be effectively discretized and can be optimized by the graph cut algorithm. Experiments on the public dataset show that the proposed model is rational and effective.