Material image segmentation based on feature similarity and non-spherical clustering

Abstract

Abstract The microstructure of the material determines its physical and chemical properties. Although image segmentation and target extraction algorithms have advanced rapidly in recent years, few segmentation algorithms for material images have emerged. By analyzing the characteristics of material images, A new image segmentation method based on feature similarity and non-spherical clustering is proposed in this paper. First, the central pixels of the region are selected according to the feature similarity, and secondly, the attribution of the remaining pixels or the non-region central points are determined by the non-spherical clustering method. In this paper, ceramic images are used as experimental materials. The results show that the proposed method achieves satisfactory segmentation effect and provides some help for subsequent image-based material analysis.