Research On Image Segmentation Algorithm Based On Improved Cuckoo Optimization

Image segmentation is a technology that dividing images into different regions and extracting useful targets,it is a key step in image analysis and is widely used in communication,military,remote sensing image analysis,and many other fields.As an important and widely used algorithm,the main process of image segmentation based on clustering is to use each pixel as a data point,and then the objective function values of these data points are calculated to make these data points be divided into different clusters.In this thesis,the image segmentation algorithm based on clustering is studied,meanwhile the Fuzzy C-means Clustering(FCM)is also analyzed emphatically.The traditional FCM algorithm is affected by the initialization of the seed points and which is easy to fall into the local optimal value.Therefore,the intelligent optimization algorithm is introduced to improve the iterative process of traditional FCM.This thesis mainly uses the cuckoo search optimization algorithm(CS)to improve the performance of the FCM.By optimizing the clustering of FCM,we put forward an extended fuzzy clustering image segmentation algorithm based on cuckoo optimization(ICS_FCM).The experimental results show that the proposed algorithm is superior to the simulated annealing fuzzy clustering image segmentation algorithm(SA_FCM)in terms of the segmentation effects and the time complexity.On the basis,drawing attention on the unsatisfactory problem of image segmentation caused by single segmentation algorithm on the process of image segmentation,this thesis transforms the problem of image segmentation into the problem of multi-object optimization by selecting two fitness functions to construct multiple object,one is based on the Euclidean distance and the other one is to merge the image pixel spatial information.Finally,a multi-object image segmentation algorithm based on improved cuckoo optimization(MOICS_FCM)is proposed.Experimental results show that MOICS_FCM owns obvious advantages in terms of segmentation effects,convergence as well as clustering validity than FCM and ICS_FCM.