Gradient Vector Flow over Manifold for Active Contours

Abstract

The gradient vector flow (GVF) snake shows high performance at concavity convergence and initialization insensitivity, but the two components of GVF field are treated isolatedly during diffusion, this leads to the failure of GVF snake at weak edge preserving and deep and narrow concavity convergence. In this study, a novel external force for active contours named gradient vector flow over manifold (GVFOM) is proposed that couples the two components during diffusion by generalizing the Laplacian operator from flat space to manifold. The specific operator is Beltrami operator. This proposed GVFOM snake has been assessed on synthetic and real images; experimental results show that the GVFOM snake behaves similarly to the GVF snake in terms of capture range enlarging, initialization insensitivity, while provides much better results than GVF snake for weak edge preserving, objects separation, narrow and deep concavity convergence.