Snakes, shapes, and gradient vector flow

TLDR

Snakes (active contours) are widely used in computer vision to locate object boundaries, but initialization issues and poor convergence to concave boundaries limit their effectiveness. The study introduces a new external force for active contours that largely resolves initialization and concavity convergence issues. The proposed gradient vector flow (GVF) force is obtained by diffusing gradient vectors of an edge map, and unlike conventional snake forces it is not a negative gradient of a potential, leading to a snake defined by a force balance condition rather than a variational formulation. Experiments on 2‑D and 3‑D examples demonstrate that GVF has a large capture range and can guide snakes into boundary concavities.

Abstract

Snakes, or active contours, are used extensively in computer vision and image processing applications, particularly to locate object boundaries. Problems associated with initialization and poor convergence to boundary concavities, however, have limited their utility. This paper presents a new external force for active contours, largely solving both problems. This external force, which we call gradient vector flow (GVF), is computed as a diffusion of the gradient vectors of a gray-level or binary edge map derived from the image. It differs fundamentally from traditional snake external forces in that it cannot be written as the negative gradient of a potential function, and the corresponding snake is formulated directly from a force balance condition rather than a variational formulation. Using several two-dimensional (2-D) examples and one three-dimensional (3-D) example, we show that GVF has a large capture range and is able to move snakes into boundary concavities.