The Research On Image Segmentation Algorithm Based On Fuzzy C-means Clustering

The Fuzzy c-means clustering algorithm could effectively deal with the problem of image blur and distortion in image segmentation.Also,it makes the image segmentation be automated without human's intervention in the process of clustering.However,the initial cluster center is uncertain,which results in setting the number of cluster categories manually,the iterative process is easy to fall into local optimum,and the spatial information of the image is not fully utilized by using the traditional fuzzy c-means clustering algorithm in image segmentation.To solve these problems,an improved fuzzy c-means clustering algorithm is proposed and applied in image segmentation.We also present the method of segmentation effectiveness.Comparing with SPFCM algorithm,our method has more advantages in noise processing and segmentation accuracy.The contents are as follows:In order to reduce the dependence of the clustering on the subjective experience and the possibility of the algorithm being trapped in the local minimum in the iterative process,a new method for selecting the midpoint of the pixel region is proposed.Firstly,the midpoint of the distance between two pixels is selected as a clustering center in the high density region of pixels.Then get out of the area.Finally recursively select the second clustering center at a distance of the farthest region using the same method.Our method makes the initial clustering center more accurate and reasonable,which has a positive impact on the clustering results.A new distance matrix combined with the gray level fluctuation is proposed to determine the pixels using the regression idea.In this thesis,we use the spatial distance and gray value information of image pixels to determine the number of clusters.According to the spatial distance of image pixel and the change of gray value,the gray level fluctuation is used as its threshold.When the fluctuation of the gray value is in the given range,a new distance matrix is generated according to the idea of stepwise regression,which is a clustering category.In our algorithm,the new distance matrix is very convenient in the storage of the intermediate results.Our method makes full use of the spatial characteristics of the image,so that the number of clusters can be determined automatically.In the past,people mainly focus on the research on algorithm improvement and ignore the method of effective determination of algorithm.On one hand,we illustrate the effect of segmentation accuracy according to the subjective impression on image segmentation results and pixel dividing accuracy.On the other hand,in order to test the segmentation affect,we propose an efficient function,which is composed of overlapping degree and separation degree to determine the best clustering result on the basis of previous studies.The experimental results show that the proposed method is effective in the segmentation results.