Contents I. Introduction
Edge is the local parts of the image whose intensity changes prominently, it reflects the image's most basic features which differ one object from another objects and is an important foundation of image segmentation, pattern recognition, image analysis and understanding. For a long time, people have mainly dedicated to the edge detection research of gray-scale image, but color image can provide more information than the gray-scale image. Studies have shown that about 90% edges of the color image are as the same as those of the gray-scale image, while the other 10% edges of the gray image can't be detected[1], [8] Thus, processing of the color image has attracted more and more attention. At present, most color image edge detection methods are realized in RGB space, that is, first gray image edge detection methods are used to three-component R, G, B, and then some logic algorithm is used to combine the edge of the three-component to get the color image edge. But the main disadvantage of the RGB space is that perceptive attribution of the color, such as hue, saturation and intensity can not be directly estimated from values of the three-component. In order to overcome the inhomogeneous and non-intuitivism of the RGB space, the color image can be processed in HSI space which better corresponds to human's visual characteristics.
