Neighborhood Space Constraint Simplifies NCM Image Segmentation

Image segmentation refers to the process of extracting different homogeneous regions in the image by analyzing and processing the image.According to the classification of segmentation strategies,the existing image segmentation methods can be divided into region-based segmentation and edge-based segmentation,neither of which can simultaneously segment certain region,edge and noise information.In order to show more useful information in image segmentation results,based on the study of Neutrosophic c-means clustering(NCM)algorithm and its simplified algorithm,Feature Space Neighborhood Constraints based on Neutrosophic c-means(FSNCS＿NCM)and Image Space Constraint Simplified Neutrosophic C-means(ISCS＿NCM)image segmentation algorithm are proposed.The specific content is as follows,(1)In order to show both edge category and noise category in image segmentation results,an image segmentation algorithm based on feature space neighborhood constrained simplified NCM clustering is proposed.Firstly,the neighborhood constraint term is constructed by combining the Euclidean distance between the mean of the measure vector of the neighborhood pixel spectrum and the center vector of the cluster.Meanwhile,the difference between the measure vector of the central pixel spectrum and the median vector of the neighborhood pixel is taken as the noise constraint term.Then the constraint terms are introduced into the simplified NCM objective function,and finally the objective function is solved.This method can not only segment the deterministic category pixels,but also effectively segment the edge and noise pixels in the image,so as to display the details of the image richly and get the segmentation results more consistent with human vision.(2)In order to improve the segmentation ability of the simplified NCM algorithm for image edge categories and the anti-noise performance of the algorithm,a simplified NCM image segmentation algorithm with image space constraints is proposed.Combining the neighborhood classification label information of pixels in the image space,define the category of the eight pixels neighborhood tag in most times and times more than average clustering center vector clustering center as a new edge.Then,according to the Lagrange multiplier method,the spectral information of the image feature space is substituted into the calculation,get the unique edge membership value corresponding to each pixel.Using the neighborhood relation of image space to achieve accurate segmentation of image edge and improve the algorithm to suppress the noise in the image.The paper include 27 figures,6 tables and 94 references.