Genetic algorithms for active contour optimization
Leonard MacEachern, T. Manku · 2002
Genetic Algorithm Snakes (GA-Snakes) are introduced as a new method of contour-based segmentation. The new algorithm exhibits several key features: a low order of complexity, the ability to handle arbitrary constraints, operation in discrete space, avoidance of higher-order derivatives, possible parallel computation, low/fixed storage requirements, the ability to handle large search spaces comfortably, and the ability to escape local minima and to handle non-convex search spaces. The resulting algorithmic complexity is of order O(n/spl lambda/G), where n is the number of control points for the contour, /spl lambda/ is the number of individuals per generation, and G is the number of generations considered.