Research On Segmentation Algorithm Based On Neutrosophic C-means Clustering

With the rapid development of information science and technology,the amount of information is increasing explosively.Image is an important way to obtain information.In general,not all information is required in the image information acquired by people.Therefore,we need to use image segmentation technology to extract and analyze the key information in the image.Image segmentation is one of the key steps in image processing,which is the necessary precondition and basis of image description and representation.In the theoretical research and practical application,it has been paid more and more attention,such as medical,military and other industries.At present,in the existing image segmentation algorithms,fuzzy C-mean clustering algorithm(FCM)is one of the most popular image segmentation methods.The most widely used algorithm,known as the basis of all algorithms.However,the algorithm does not take into account the relationship between pixels,vulnerable to outliers and the classification effect is not obvious.Many scholars have carried out a series of development and improvement on the basis of this method,which makes the method more effective.Therefore,Neutrosophic C-means clustering algorithm(NCM)is proposed,the algorithm is inspired by the FCM algorithm and neutrosophic set framework,although FCM is shortened long calculation time,improve the classification accuracy,but poor anti noise has not been improved.In order to enhance the anti noise performance and robustness of NCM method,the 2D histogram,neighborhood spatial information,the regeneration of Hilbert kernel function concept into the algorithm,improvement and optimization,put forward a series of new algorithms,and carried out the following research contents on the several algorithms:1.The fuzzy set theory,FCM and NCM are introduced to analyze the similarities and differences between FCM and NCM.A new algorithm is proposed based on the combination of the pixel values of the neighborhood pixels,the combination of each pixel and the neighborhood of the neighborhood pixels to form a binary array,and the frequency of its occurrence.This paper proposes a new algorithm based on two-dimensional histogram Fuzzy-C-means clustering algorithm(Neutralophic C-means Clustering Algorithm Based on Two Dimensional Histogram,2DH-NCM).In the image segmentation visual effect,the new algorithm has better anti-noise ability and segmentation effect than the fuzzy C-means clustering method.The peak signal to noise ratio(PSNR)of the segmented image shows that the PSNR value of the new algorithm is larger than that of other algorithms,which is 5-7dB higher than that of the FCM algorithm,which is larger than that of the NCM Algorithm 2-3dB.2.NCM algorithm has poor clustering performance for non-convex irregular data.Therefore,the concept of regenerated Hilbert's kernel function is embedded in the NCM algorithm,and the kernel-induced distance is used instead of the Euclidean distance.Using the non-linear problem of Mercer condition,the low-dimensional linear indivisible model is mapped to the separable high-dimensional space,Sample characteristics.The results show that the algorithm is more stable and the anti-noise ability of the algorithm is improved by using the NCM algorithm(Neutrosophic C-means Clustering Algorithm in Kernel Space,KNCM),which is a new algorithm.3.In order to improve the robustness of the NCM algorithm,the neighborhood information constraint function is embedded in the algorithm,and the pixel is fully correlated with the algorithm,and the neighbor information constraint NCM algorithm(Neutralophic C-means Clustering Algorithm in Location Information,LNCM)is proposed.Through the image denoising test,it is proved that the new algorithm can meet the image segmentation requirement and has good robustness.4.The neighborhood information function and the regenerated Hilbert kernel function are embedded into the objective function of the NCM algorithm.The relationship between membership degree and clustering center is established,and the kernel distance is used instead of Euclidean distance to optimize the sample characteristics.The kernel function NCM algorithm for obtaining local neighborhood information.Compared with other algorithms,it is proved that the segmentation result is stable,the boundary is smoother,suitable for image segmentation,and has high robustness and noise immunity.