Invariant Topology Snakes Driven by Particle Swarm Optimizer
Delu Zeng, Zhiheng Zhou · 2008
Active contour modelling (ACM) plays a significant role in boundaries extraction or image segmentation. In this paper, we present a method of invariant topology snakes driven by particle swarm optimizer (ITPSO snakes) and its application in ACM. We use some control points on the contour as particles in ITPSO. To avoid intersections and maintain topology of the active contour in the process of ACM, and consider inhomogeneous boundaries, we design a novel scheme to restrict the velocity updates. So the best discovered positions for next round velocity updates are no more computed as in the traditional method of PSO. The experimental results show that the proposed ITPSO snakes can perform well in homogeneous boundaries extraction.