The Image Segmentation Methods Research Based On Clustering Technology

In image processing, image segmentation is a fundamental problem, based on color, text, shape and spatial location features, image is divided into meaningful parts.From the Angle of theory, image segmentation is designed to find the data subset natural together of the data set, as is known to all, this is a clustering problem.Due to the binary image plays an important part in image processing, this article studied separately the binarization segmentation of image, first put forward using k medoids algorithm to process the image into binarization image.When k = 2, the image be processing into binary image naturally.Among various kinds of clustering technology for image segmentation, the spectrum method is a kind of general method. But in traditional spectral clustering method, the final stage of clustering using the k-means clustering algorithm for cluster operation, due to defects of the k-means algorithm is sensitive to the initialization,most likely to converge to the local optimum Makes when the traditional spectral clustering method is applied to image segmentation, the image’s segmentation effect is not very good.In order to solve the defects of the traditional spectral clustering method in the k-means clustering stage, this paper introduces the genetic algorithm to the spectral clustering algorithms, using genetic algorithm(ga) to replace the k-means clustering algorithm of spectral clustering algorithm, propose genetic spectral clustering algorithm, the proposed genetic spectral clustering algorithm greatly improve the performance of spectral clustering algorithm and clustering quality.the comparison to the traditional spectral clustering on UCI data sets, a synthetic data set,and real image data sets and, the result of the experiment proves the superior performance and clustering effect about the proposed genetic spectral clustering algorithm.In this paper, the main research work are as follows:1. The k center clustering algorithm is applied to the image binarization processing, proposed the IBKMC algorithm and FIBKMC algorithm for the binarization option of gray image.2. To improve the traditional spectral clustering algorithm, combining spectral clustering algorithm and genetic algorithm to form genetic spectral clustering algorithm(GSC).3. As mentioned in 2th point, genetic spectral clustering algorithm is applied to the study of image segmentation.4. Studied the method of image segmentation, the theory of clustering technology, the theory of genetic algorithm, provides a solid theoretical foundation for latest studies.