Abstract:

Image segmentation mainly extracts the objectives users are interested in, and it is the basis for image classification and pattern recognition. We present a novel image segmentation method based on fractional-order Darwinian particle swarm optimization, called FODPSO   . The algorithm utilizes the fractional calculus strategy to control the convergence of particles and is able to determine the n-1 optimal for n-level threshold on a given image. Compared with the APSO and the CFPSO algorithms, testing results show that the FODPSO algorithm can enhance the performance in terms of convergence speed, stability, solution accuracy and global optimality, and greatly overcome the shortcomings of traditional methods, such as local optima and slow convergence speed. Hence, the FODPSO is applicable to practical projects.

Key words: multi-scale segmentation, fractional-order Darwinian particle swarm algorithm, class variance, algorithm comparison