Application of grid-based C-means clustering algorithm for image segmentation

C-means clustering algorithms have proven effective for image segmentation, but are limited by the following aspects: 1) the determination of a priori number of clusters. If the number of clusters can be incorrectly determined, a good-quality segmented image cannot be assured; 2) the poor real-time performances due to great time-consuming, and 3) the poor typicality of each cluster represented by the clustering prototype. In this paper, a grid-based C-means algorithm is applied to image segmentation, whose advantages over the existing C-means algorithm have demonstrated in some typical datasets. The convergence domain of the grid-based C-means algorithm has further been analyzed as well. Experiments show that the grid-based C-means algorithm outperforms the original C-means algorithm in some typical image segmentation applications.