Image segmentation algorithm based on dynamic particle swarm optimization and K-means clustering

In order to solve excessive independence of image segmentation quality of K-means clustering algorithm on initial clustering center for selection, and easily falling into the local optimal solution etc., one kind of image segmentation algorithm, dynamic particle swarm optimization and K-means (DPSOK) based on dynamic particle swarm optimization (DPSO) and K-means clustering was proposed in the Thesis. The performance of PSO algorithm was strengthened by dynamically adjusting inertia coefficient and learning factor; then fitness variance of particle swarm was calculated, and opportunity to transfer to K-means algorithm was found accurately; then K-means clustering center was initialized by utilizing DPSO output result to make it converge to the global optimal solution. Finally, K-means clustering center was continuously updated by minimizing multiple iterations of the target function until convergence. It is shown in the experimental result that DPSOK can effectively improve the global search capacity of K-means, and it has better segmentation effect compared with K-means and PSO in image segmentation. Compared with particle swarm optimization and K-means (PSOK) algorithm, DPSOK algorithm in the Thesis has higher segmentation quality and efficiency.