Polygonal Approximation of Digital Curves Using the State-of-the-art Metaheuristics
Peng-Yeng Yin · 2007
In this chapter, we investigate the polygonal approximation problem which is fundamental to many image analysis tasks. Traditional problem-specific heuristics are not suitable to be applied alone because the quality of the obtained result depends on the initial setting of the algorithms and the properties of the curves. On the other hand, metaheuristic approaches can produce stable approximation quality for various kinds of curves. We have illustrated the implementations based on two newly developed metaheuristics, namely the ACO and the PSO. To circumvent the underlying problem, specific features have been introduced such as the ACO graph representation, PSO genetic operators, penalty functions, and the hybrid strategy. Experimental results on several benchmark curves have manifested that these new features can improve the performance of metaheuristics in solving the polygonal approximation problem.