Studies On Color Space Selection And Segmentation Quality Evaluation Methods

Image segmentation is a key technology of image processing, and its purpose is to separate the target from the background. It makes further image analysis and comprehension possible. But so far, there is not a unified evaluation criterion about the effects of image segmentation.Segmentation method is the most important factor of the segmentation result, so the domestic and foreign research focused on improving and optimizing the segmentation methods at present, and the color space used in image segmentation also affects the segmentation result in certain extent, segment the image in the appropriate color space, the segmentation result will be better. In response to this question, five representative color spaces(RGB,HSI,YIQ,YCbCr,Lab) and their principal component color space(PRIN) are compared in this paper. A large number of experiment shows that PRIN color space is the best choice for image segmentation, but it requires a lot of computation, RGB color space behind PRIN space, and then YIQ and YCbCr color space, HSI color space is worst.At present, although a number of different types of segmentation evaluation methods have been proposed, but most evaluation methods are proposed for specific applications.A segmentation quality evaluation method based on the number of pixels which is categorized improperly is proposed in this paper, and this paper compared the segmentation image of each color space. The study implies that the evaluation results by the method of this paper is consistent with the final experimental results, it can judge the results of the segmentation effectively. In addition, because of the correct rate of segmentation is used in this evaluation method, it not only can compare a number of segmentation results, but also can be used to evaluate the separate segemetation problem, and the scope of application is broad.