Improved Gradient Vector Flow over Manifold for Image Segmentation Using Active Contours
Xiaoteng Zhu, Ke Cheng, Qingfang Chen, Yuanquan Wang, Ziyang Zhang · 2021
In this article, an improved gradient vector on manifold (IGVFOM) model was proposed for active contours. In this proposed IGVFOM model, two components and gradient of image coupled together during the diffusion by generalizing the Laplacian operator from flat space to manifold. Since coupling terms could fuse information of image during diffusion, the IGVFOM snake behaves better than GVF snake on weak edge preserving, noise robustness, objects separation, narrow and deep concavity convergence while keep advantages in terms of capture range enlarging and initialization insensitivity. Experiments on synthetic and real images manifest these properties of the IGVFOM snake.