Fuzzy Clustering And Image Segmentation Algorithm Based On Generalized Equalization

Image segmentation refers to each image into a series of mutually overlapping homogeneous regions.It is one of the basic problems of image processing and computer vision is to achieve from the image processing to image analysis,image understanding and then completed a critical step,and has been widely in practice,such as computer vision,pattern recognition and medical image processing,etc.Applications.Various image segmentation methods applied today,the fuzzy C-means clustering algorithm is widely used because of the use of iterative methods to achieve the classification of samples.However,the FCM algorithm does not consider the impact of different types of samples on the classification decision,some scholars have improved the objective function of FCM,and put forward the generalized fuzzy C-mean clustering algorithm.The information of sample size is used in the objective function,so as to reduce the effect of different sample number on the classification result.Researchers have also established non membership parameters,introduced the "betwixt the fuzzy concept of traditional fuzzy set to optimize.The FCM algorithm is extended to intuitionistic fuzzy sets and intuitionistic fuzzy C-means clustering(IFCM)algorithm is obtained.This paper is mainly based on the existing fuzzy clustering segmentation methods,in-depth study of the GEFCM and IFCM methods,analysis of the existence of these two kinds of clustering methods of the problem;and through the analysis and test of the improved algorithm.The main work of this paper is as follows:1.Aiming at the shortcoming of existing general equalization fuzzy C-means clustering algorithms that are not convergent.A new general equalization fuzzy C-means clustering algorithm is proposed and its convergence and classification performance are deeply studied.On the basis of the objective function of existing general equalization fuzzy C-means clustering,the properties of the limit expression of Schweizer T-norm is used to construct the objective function of a new general equalization fuzzy C-means clustering,the Lagrange multiply method is adopted to obtain iterated formulas of the fuzzy membership and clustering center for the modified general equalization fuzzy C-means clustering and the convergent problem of this clustering algorithm is proved by Zangwill theorem.In order to further improve the performance of the clustering algorithm,the iterative expression of the clustering center is modified and the modified clustering algorithm of significant improvement of a class of clustering performance is achieved.The experimental results of clustering analysis of Iris data and gray image segmentation indicate that the proposed general equalization fuzzy C-means clustering algorithm is efficient and its modified algorithm can obtain more satisfied clustering quality and segmentation effects than existing different fuzzy c-means clustering algorithms.2.when using the modified general equalization fuzzy C-means clustering algorithm to cluster analysis,first class distribution is assumed hypersphere or super-ellipsoid,when all kinds of boundary sample is linearly inseparable or class distribution is not ultra-sphere or hyperelliptic when the body is often the case clustering failure or misclassified appears.This approach did not feature samples were optimized,but the direct use of clustering feature samples,the effectiveness of the algorithm depends largely on the distribution of the sample,it can be introduced into the kernel-based learning method is proposed based on kernel spaces general equalization fuzzy C-means clustering algorithm to increase the optimization of sample characteristics.Through the use of nuclear function in the observation space linearly inseparable samples nonlinear mapping to high-dimensional feature space becomes linearly separable,and sample characteristics by good resolution,extracted and amplified,can achieve more accurate clustering.3.In order to improve the effectiveness and robustness of the intuitionistic fuzzy C-means clustering in image segmentation,a new segmentation method of intuitionistic fuzzy C-means clustering in Kernel Space with local information constraint is proposed.In view of the shortcomings of the existing intuitionistic fuzzy C-means clustering,use nonlinear function to map data samples from the Euclidean space to the high dimensional feature space of Hilbert,The intuitionistic fuzzy clustering algorithm in Kernel Space is obtained by using intuitionistic fuzzy clustering.At the same time,it is applied to the image segmentation.And taking into account the interaction of neighboring pixels,the neighborhood pixels are integrated into optimization of objective function of intuitionistic fuzzy clustering in Kernel Space Algorithm,intuitionistic fuzzy clustering segmentation in Kernel Space with the local information of pixels is obtained by mathematical deduction.The tests of Gray Graph Segmentation show that compared to the existing intuitionistic fuzzy C-means clustering segmentation method,intuitionistic fuzzy C-means clustering algorithm in Kernel Space is more satisfactory in the results of segmentation.And the Kernel Space of intuitionistic fuzzy C-means segmentation method with local information is shown to be more robust.