Intuitionistic Fuzzy Suppressed FCM Image Segmentation Based On Spatial Information

Image segmentation is a key step in the field of image processing and computer vision,and it is also a challenging research hotspot.The fuzzy C-means(FCM)clustering algorithm is widely used because of its ability to handle the fuzziness of images among many image segmentation algorithms.However,fuzzy sets have some limitations in describing the fuzziness of images.Intuitionistic fuzzy sets are generalized concepts of fuzzy sets,so it has a stronger ability to deal with the fuzziness of things.In this paper,the intuitionistic fuzzy sets and spatial information are taken as the main theoretical tools to solve the problems of the FCM algorithm and its improved algorithms: noise sensitivity,low convergence speed,etc.The specific work is as follows:(1)Suppressed fuzzy c-means(SFCM)algorithm is an improved algorithm of FCM algorithm,which solves the problem of low convergence speed of FCM algorithm.However,the SFCM algorithm inherits the defect that is sensitive to noise and generates a new problem: that is,how to realize the selection of the inhibitory factor adaptively.Therefore,this paper proposes a suppressed non-local spatial intuitionistic fuzzy c-means(SNLS-IFCM)image segmentation algorithm.The algorithm considers the non-local spatial information of the pixel,and generates the hesitation degree as the inhibitory factor according to the “voting model” in the intuitionistic fuzzy set to realize the adaptive selection of the parameter.The experimental results show that the SNLS-IFCM algorithm has a better segmentation effect on noisy images and the operation efficiency is improved.(2)In order to solve the issues that the intuitionistic fuzzy c-means(IFCM)algorithm does not consider the spatial information of pixels and slow convergence,a kernel spatial adaptive suppressed intuitionistic fuzzy c-means(KSAS-IFCM)image segmentation algorithm is proposed.The algorithm considers the neighborhood mean spatial information of the pixel,and constructs the intuitionistic fuzzy membership degree by using the hesitation degree generated by the “voting model” to reduces the influence of artificial parameters on the experiment.Secondly,an adaptive selection formula of the inhibitory factor is given according to the gray feature and spatial distance of the pixel.Finally,the distance between the pixel and the cluster center is calculated using the kernel-induced distance.The experimental results show that the KSAS-IFCM algorithm has a better segmentation effect on noisy images.However the algorithm takes a long time to run due to the increase of computational complexity.(3)To some extent,the local spatial information and the non-local spatial information are complementary.This paper integrates the neighborhood median spatial information of the pixel based on the first improved algorithm,and a kernel complementary spatial suppressed intuitionistic fuzzy.c-means(KCSS-IFCM)image segmentation algorithm is proposed.The algorithm also incorporates the characteristics of the second improved algorithm,which uses the kernel-induced distance instead of the eucildean distance,and uses the hesitation degree and membership degree to construct the intuitionistic membership degree.The experimental results show that the KCSS algorithm can achieve better segmentation results under the interference of gaussian noise and salt&pepper noise.