A New Algorithm Of Image Segmentation Based On Pixel-level Multi-feature

Image segmentation is an important tool for image processing, can be used as an effective front-end complex algorithms, which simplifies subsequent processing. This low-level visual tasks is often the initial, but also essential have applications on the video, and computer vision, such as targeting or recognition, data compression, tracking, image retrieval and so on. Although a large number of segmentation algorithms on the majority of researchers have been proposed and improved, but no method is perfect, suitable for any kind of image segmentation, there is a lack of methods or algorithms. So far, image segmentation is an important tool, at the same time, it is still a challenging problem in the field of image processing.Currently, based on pixel-level ,variety of features segmentation algorithm and a variety of Combination of segmentation has become a hot research field of image segmentation. Through careful study and summarize, this paper is superior to the traditional fuzzy clustering algorithm and support vector machine theory and experimental results were compared and analyzed cf comprehensive system to pixel-based color, texture and other characteristics of digital images based on division of research, this paper though fuzzy clustering algorithm and support vector machines research and analysis, complete the following tasks:1. The paper detailed in-depth study in particular, fuzzy clustering algorithm fuzzy C means clustering (FCM, Fuzzy C-Means) segmentation algorithm, carefully study the image segmentation algorithm based on fuzzy clustering in the number of categories to determine the initial cluster, The initial cluster centers and membership functions of choice.2. The fuzzy C means clustering (FCM, Fuzzy C-Means) theory is a combination of texture measurement and adaptive laws threshold image segmentation algorithm FCM. Through the experiment, simulation results show that the adaptive threshold method of image segmentation results human visual perception system a good match, can effectively suppress background noise while saving run-time image segmentation to improve the speed of image segmentation.3. Through some factors about support vector machine (SVM) ,such as the type of kernel function, kernel parameters, the penalty factor and so on,on the methods for image segmentation function analysis, research, and presented based on unsupervised support vector machines for classification, the application Support Vector Machine (SVM) provides a basis for image segmentation. 4. Through the study of a variety of image segmentation based on the proposed pixel-based color image segmentation algorithm, respectively, using homogeneity, Gabor filters to extract the color and texture features, better than the traditional use of support vector machine learning Machine (SVM) classification. Experimental results show that the method itself made the experiment results better and more effective than before, and made the time lower, and recently in the literature compared the proposed method improves the quality of color image segmentation.