Color Image Segmentation Based On Fuzzy Clustering Method

Image segmentation is the technologies and processes of decomposing the image into the some areas each with their own characteristics and also extracting the objectives of interest. Image segmentation is a very important and difficult problem in the fields of image understanding, pattern recognition and computer vision. It's also the first critical steps in the computer vision technology. With the development of technology, color image has got more and more attention. Compared with gray image, color image has more information. It not only contains the brightness information, but also the hue, saturation, etc.Color image segmentation has two main aspects. The first is to choose the appropriate color space; the second is to use appropriate division strategies and methods. The common color space is RGB space. when we split out the color image-we often can't get the results that we need, because the r, g, b three-components'relationship is very high (For example, the area with different brightness and with the same color will be divided into different regions.) HIS space is non-linear transformation of RGB space, It removes the correlation between the three component. This space is more in accord with human visual characteristics, because color information H, S and brightness information have been separated. When we only need tonal segmentation, we can use hue H component alone to split., So I choose HIS color space.Researchers continue to improve the original method of image segmentation, and some new theories and new methods of other disciplines are used for image segmentation, meanwhile, they made a lot of new segmentation methods. The existing color image segmentation methods mainly include Threshold-based segmentation method, Region-based segmentation method, Fuzzy method, the neural network method, the clustering method, Edge-based segmentation method, Based on the theory of particular segmentation method and so on.In the image acquisition process, the image is often mixed with noises which make the image to blur deformation. Fuzzy theory has the ability to describe the uncertainty problems. This article focuses on the fuzzy clustering method to study on color image segmentation problem. Traditional FCM clustering algorithm can't effectively separate the noise image. This article will introduce the spatial information of pixels and combine with the color information to execute the fuzzy C-means clustering. I discuss the impact of each parameter, and also through experiment shows the method has good noise removal effect. In the practical application of checking the number of cigarettes, I can well split out the cigarette butts, meanwhile, I can get the exact number of cigarettes, however, Its background has no effect on segmentation.