The Study Of Support Vector Machine For Color Image Segmentation

Image segmentation is an important image analysis technology, which refers to the process and technology that divides the image into regions whose local features (texture, chrominance, luminance, etc.) are consistent with each other, and then extract the target which human are interested in. It is not only the crucial step in which the image processing is transferred into image analysis, but also an important research field of computer vision. With the development of computer technology, as color images can provide more abundant information than the gray image, color image application and studies deepens. Image segmentation is also increasingly affected by the attention of scholars, and the corresponding algorithm is endless.In recent years, support vector machines, as a new method of pattern recognition, applies for pattern recognition, regression analysis and probability density estimation and other aspects, Compared with the traditional unsupervised classification method, it is more suitable for actual field application, and more efficient and simpler.This paper discuss support vector machine theory from the view of reality and applies it for image segmentation, there are two main aspects:(1) Edge detection segments images by detecting edges among different areas. For color image edge detection, edge information of luminance and chrominance channel of color images is analyzed first, then for the features of the color image edges, multi-dimensional vectors are established. The trained support vector machines have the ability to identify specific target edge, overcome shortcomings of the traditional edge detection algorithm such as being non-targeted and noisy.(2) By analyzing the supervised and unsupervised learning methods in color images, color image segmentation model consisting of two methods is presented. This model combines the advantages of both methods, having self-support capabilities to realize the adaptive segmentation of color images as well as good versatility and real-time feature. Experimental results show that the model can meet the color image segmentation requirements in a lot of visual applications.