Image Segmentation Method Based On Fuzzy Clustering Algorithm Is Studied

Image Segmentation is an important step in Image Engineering which got lots of attention from research community at the beginning of image engineering. Image segmentation technology is used in each level of image engineering fields; it used from the bottom of the pixel-level processing to the image of the middle layer and advanced applications of image understanding. The application of the breadth and depth is always the important cause of prosperity.Image segmentation algorithm, the cluster analysis plays a very important role. It derived from mathematical set theory and it is an important tool for data analysis. One of the most classic methods which from many clustering analysis method is the Fuzzy C-means clustering algorithm (Fuzzy C-Means Clustering FCM). FCM is a representative of the vagueness and uncertainty theory which has been widely used in academic and social life in all fields and application level. It is belongs to unsupervised clustering algorithms, which has advantages of the strong convergence and describe things fuzzy, but in fcat, FCM also has shortcomes: dependence on the initial value; speed on large data model too slow; stringent signal to noise ratio; easy convergence in local minima. In this paper, we put forward the improvement from the characteristics of the FCM algorithm, the main work is as follows:1. FCM algorithm runs too speed in large data space. In this paper, we presents an improved algorithm which based on quadtree (QT-FCM). We used the quadtree segmentation algorithm for reducing the data space of the traditional FCM algorithm,and then convert the traditional elements of the data space into a collection space.2. FCM segmentation algorithm based on the histogram got a narrowed data space that can help it got greater improvement in clustering efficiency. But it has some Applicability in large noise sensitive and considerations on neighborhood informations. Based on such considerations, in this paper, we presents three-dimensional histogram based on neighborhood information in the FCM algorithm (HFCM), by superposition neighborhood information to change the data space, and thus efficient use of the image of the neighborhood information. The process is as follows:Using the noise of the original image and filtering the images to combinate a three-dimensional gray-scale histogram which has the two-dimensional statistics to streamline Statistical data. Experiments show that the algorithm has a slight increase in the calculation and the high signal to noise ratio of the image. It can achieve better clustering effect.3. Segmentation algorithm for the membership function prototype for the FCM analysis. In this paper, we present an improved FCM algorithm which based on the thought of the minimum rivals (SRC-FCM). The way is weaken the weakest membership and increasing the maximum ones’,and then, It can improve the moving speed of the cluster center, and thus improve the FCM algorithmoperating efficiency. The theoretical analysis and experiments have proved that this method has faster convergence than the prototype.The algorithm is applied to the large noise in the image, and achieved good segmentation results, increase the segmentation speed.The experiments verify the effectiveness of the improved algorithm.