Research On Robust Fuzzy Clustering Segmentation Algorithm Based On Master-slave Hierarchical Local Information Constraints

The essence of image segmentation is to divide the image pixels into several different regions,and ensures that the pixels in the same region have the similar feature information,while different regions have obvious differences.Among many image segmentation methods,clustering-based segmentation method has been widely applied.However,the uncertainty of the image itself or the incompleteness of information and the ambiguity of segmentation may affect the effective segmentation of image information.Therefore,it is an inevitable trend to introduce fuzzy theory into image segmentation.Fuzzy C-Means(FCM)clustering algorithm based on fuzzy set theory is simple in iterative process,low in computational cost and easy to implement.FCM does not make use of the neighborhood information of samples,and has the constraint that the sum of membership degree of sample to all clusters is equal to 1.Therefore,noise and outliers have great influence on the performance of the algorithm,and the robustness of the algorithm against noise is very poor.The effective way for experts and researchers to improve FCM is to add local neighborhood information and possibilistic typicality to the objective function of clustering algorithm to enhance the anti-noise robustness of the algorithm.Based on this,this paper uses local spatial information constraints and possibilistic theory to improve the existing algorithms,and proposed a series of improved robust fuzzy clustering-related algorithms.The specific research content is divided into the following parts:(1)Local neighborhood information plays an important role in robust fuzzy clustering-related segmentation algorithms,and how to construct local information items is the key to robust fuzzy clustering.Based on existing local information constraints,this paper proposes a model to describe the hierarchy relationship of local neighborhood windows in the master-slave neighborhood model,which combines spatial distance with gray information to suppress the influence of noise on current pixel clustering,and can also well control the balance of noise suppression and detail preservation.Based on this model,this paper proposes a robust fuzzy clustering segmentation algorithm with master-slave neighborhood information constraints,and its convergence is strictly proved by Zangwill theorem and bordered Hessian matrix.Experiments results show that the proposed algorithm has good segmentation performance and strong anti-noise performance,even significantly outperforms existing state-of-the-art robust fuzzy clustering-related algorithms in the presence of high noise.(2)Kernel weighted fuzzy local information C-means clustering is of great potential significance in noisy image segmentation,but it has obvious defects for image with high noise.Considering that existing local neighborhood without rich local information,the existing robust fuzzy clustering with local information is improved by using the master-slave neighborhood model which deeply explores the local neighborhood information,and then the improved clustering algorithm is extended to the kernel space,and a novel fuzzy C-means clustering with kernel metrics and richer local information is proposed.Experiments show that the proposed algorithm is effective for noisy image segmentation,and it is obviously superior to existing robust fuzzy clustering-related algorithm,which has far-reaching significance for robust fuzzy clustering theory.(3)Considering the defect of possibilistic fuzzy clustering with weak noise-resistant robustness,a robust possibilistic fuzzy additive partition clustering driven by deep local information for segmenting images with high noise is proposed in this paper.Firstly,the proposed master-slave neighborhood information is integrated into the possibilistic fuzzy additive partition clustering model,thus a new robust possibilistic fuzzy clustering model with master-slave neighborhood information constraints is constructed,in which the master-slave neighborhood information consists of the master neighborhood window of current clustering pixel and the slave neighborhood windows around the master neighborhood pixel.Next,this model is further simplified by Cauchy inequality and a robust master-slave neighborhood information-driven possibilistic fuzzy clustering algorithm is derived by Optimization theory.Finally,using Zangwill theorem and bordered Hessian matrix,the proposed algorithm is proved to be locally convergent.Testing results indicate that the proposed algorithm is very effective for noisy image segmentation,and its segmentation performance is obviously better than many existing advanced fuzzy clustering correlation algorithms.(4)Aiming at the obvious defects of kernelized possibilistic and fuzzy additive partition clustering motivated by local information to segment noisy images,a kernel function-driven possibilistic and fuzzy clustering incorporating depth local information is proposed.Firstly,the master-slave neighborhood which can deeply explores richer local neighborhood information is introduced into the possibilistic fuzzy additive partition clustering model and an enhanced possibilistic fuzzy depth local information clustering algorithm is obtained.Then,this extended algorithm is kernelized to further strength the noise-immune robustness of the algorithm.Finally,optimization theory is used to analyze the convergence of the algorithm.Testing results indicate that the improved algorithm has good segmentation performance for noise image,significantly outperforms many robust fuzzy clustering correlation algorithms,and has far-reaching significance to facilitate the development of robust segmentation theory.