Variable-length PSO optimized fuzzy clustering for self-adaptive image segmentation

The image segmentation method based on fuzzy c-means (FCM) clustering is sensitive to the initial values, and the cluster number should be given before clustering. Therefore, an adaptive image segmentation method based on PBMF fuzzy clustering optimized by variable-length particle swarm optimization (PSO) is proposed. The PBMF index function considers the cluster number and the cluster centers. The method designs a variable-length PSO in the optimization of the PBMF index function, utilizes the statistical histogram to convert the pixel space to the gray histogram characteristic space of the image, and then quickly obtains the optimal cluster number and cluster centers. Experimental results of segmentation for a remote image indicate that the self-adaptive segmentation strategy is characterized by strong global capability of searching the optimal cluster number and cluster centers, and strong anti-noise ability.