Research On Improved Fuzzy C-Means Algorithm And Its Application In Image Segmentation

With the advent of the era of big data,information is showing explosive growth.How to extract the information we need from these massive data is particularly important.As an important way for us to understand the world and obtain information,various image processing methods have received great attention.Image segmentation is a very critical link in image processing,and clustering is one of the core methods.Due to the variety of ways to obtain information,the accuracy of the information cannot be guaranteed,and the method of image segmentation based on clustering method in fuzzy environment came into being.Among these methods,Fuzzy C-Means(FCM)algorithm stands out among lots of fuzzy clustering algorithms,and because of its simple implementation and good segmentation,it is widely used in image segmentation.However,the FCM algorithm does not contain the spatial information of pixels,which is also one of the obvious disadvantages,and it is difficult to suppress the influence of noise.Secondly,the initial value needs to be set manually,and there is randomness.In some improvements,some constraint items or spatial information are added,which is easy to cause the algorithm to take too long to run,reducing the convergence speed,resulting in inefficient algorithm.Based on the above reasons,this dissertation presents improved image segmentation methods for integrating existing spatial information,enhancing the noise resistance of the algorithm,reducing the time complexity of the algorithm,and improving the algorithm after generalizing some fuzzy sets.Six aspects are summarized.The research improves the efficiency of the algorithm while accelerating the algorithm's convergence speed while overcoming the FCM algorithm's sensitivity to noise.The improved algorithm is applied to image segmentation.The main research results obtained are as follows:(1)In order to solve the problem that FCM is easy to fall into local optimality,this dissertation proposes a new intuitionistic fuzzy entropy formula,and uses it to obtain the sample feature weight,and then adopts the greedy idea to select the initial clustering center from the points with larger feature weights.Introduce the intuitionistic fuzzy entropy formula to obtain the feature weights to obtain the feature vector.Calculate the Euclidean distance with the feature vector as the center to define the regional density.Only select the initial clustering center in the high-density area and the distance as far as possible from each other.The example of the experiment is verified by experiment to verify the validity and feasibility of the algorithm.(2)In order to make the algorithm more capable of handling fuzzy problems,this dissertation generalizes fuzzy sets to intuitionistic fuzzy sets,and extends the kernel induction distance to the intuitionistic fuzzy kernel induction distance instead of the Euclidean distance,introduces suppression factors through the idea of competitive learning,and uses hesitation generate and use to modify the membership degree,and add weighted local space and gray information constraints to modify the objective function.Then,an image segmentation experiment is performed to verify the effect of the algorithm.The segmentation objects include the original image and the original image with noise,respectively,and the experimental results are analyzed.(3)Due to the neutrosophic set has more advantages in processing fuzzy boundary information than the intuitionistic fuzzy set,this dissertation further extends the intuitionistic fuzzy set to the neutrosophic set,and uses the distance measurement formula that can be adaptively generated as the suppression factor to modify the true membership and obtain adaptive suppression In the formula,the FCM algorithm is used to verify the image segmentation experiment of the improved algorithm.The original image and the original image with noise are segmented separately,and the experimental results are analyzed.(4)The two improved algorithms proposed in this dissertation segment gaussian noise image and salt and pepper noise image respectively,and analyze and compare with the effect images obtained by FCM algorithm and IFCM algorithm segmentation to verify the superiority of the algorithm.At the same time,the evaluation index is introduced to quantitatively analyze the improvement effect of the algorithm and verify the effectiveness of the improved algorithm proposed in this dissertation.