Improvement Of K-means Clustering Algorithm And Its Application In Color Image Segmentation

Cluster analysis is the main content of data mining.K-means algorithm is one of the common clustering algorithms.K-means algorithm is widely used in many fields due to its advantages such as simple thought,easy coding,fast convergence and so on.However,K-means algorithm also has some defects.In order to better excavate high-quality information,it is urgent to study the defects of K-means algorithm.In image processing,whether image analysis or further understanding of image semantics,it is often necessary to perform image segmentation on the original image,and then perform feature extraction on different regions obtained by segmentation.In the field of image segmentation,clustering algorithms have attracted wide attention due to their advantages such as high efficiency and adaptability.This paper studies cluster analysis and color image segmentation.The main work as follows:(1)The main research contents are cluster analysis algorithms and color image segmentation algorithms.(2)Aiming at the defect that the number of clusters is difficult to determine and the clustering result is sensitive to the initial clustering center in the traditional K-means algorithm,an improved K-means algorithm(CNACS-Kmeans)is proposed.A new method for calculating the local density of the data object is defined,a decision graph for the data set is constructed,and the number of optimal initial cluster centers and clusters is obtained from the decision graph using regression analysis and residual analysis.Then use the obtained initial cluster centers as the input parameters for clustering operation.Experimental results on simulated data sets and UCI real data sets showed that the improved algorithm can achieve better clustering results.(3)It is difficult to determine the number of segments when using K-means to segment color images.In addition,there are also phenomena of over-segmentation and wrong-segmentation when using K-means to segment a color image.In order to solve these problems,a color image segmentation method based on improved K-means pre-segmentation and region merging is proposed.Firstly,the selection strategy of the initial center of K-means algorithm in segmenting color images is improved.Next,set a large K value to perform initial segmentation of the color image using the improved K-means algorithm.Next,merge the adjacent and similar regions obtained from the initial segmentation.When the change of the rate of image color dispersion exceeds the threshold after one merge,the merge should be cancelled and the final segmentation result is generated.Experimental results show that the segmentation algorithm can achieve better segmentation results.