Adaptive fractional-order genetic-particle swarm optimization Otsu algorithm for image segmentation

The two-dimensional (2D) Otsu algorithm used in image segmentation considers both the gray scale information of the image and the spatial information contained in the neighborhood of pixels. The algorithm is quite effective, but a number of variations have been proposed to improve its performance. In this paper, a new adaptive fractional-order (FO) genetic-particle swarm optimization (FOGPSO) version is proposed. The FOGPSO associates the particle selection operations of a genetic algorithm (GA) and particle swarm optimization (PSO). Crossover and genetic mutation are used to avoid PSO falling into a local optimum. Fractional calculus operators are adopted in the updating scheme of velocity and position, while the order of the derivative is adaptively changed according to the state of the particles. Compared with the original PSO, the integer PSO, the fractional PSO (FOPSO), other improved versions of the PSO for Otsu algorithms, and other existing methods for 2D Otsu algorithms, the proposed method shows great superiority. Indeed, experimental results reveal that both qualitatively and quantitatively, through suitable indices, as the regional contrast, the intersection over union (IOU) and peak signal to noise ratio (PSNR), the FOGPSO outperforms the other methods, thus verifying the effectiveness of the new algorithm.