Two-Dimensional Fuzzy Clustering Algorithm (2DFCM) for Metallographic Image Segmentation Based on Spatial Information

Image segmentation has a positive impact in materials science, and it has application prospect and research value especially in the forecast of material performance. Considering spatial neighbourhood information can improve the accuracy of image segmentation, a novel modified FCM method for image segmentation is presented in this paper. This method take full advantage of the relevance of the current pixel to its neighbour pixels, then design a simple and effective two-dimensional distance metric function, and build a new objective function, so that the cluster centers are updated simultaneously in two dimensions of the pixel value and its neighbouring value. Experiment results of metallographic image segmentation showed that the algorithm has better noise immunity and better convergence rate than the conventional FCM algorithm.