A NOVEL DENSITY GRADIENT-BASED COLOR EDGE DETECTION APPROACH

The intensity gradient-based methods are commonly used in edge detection of images. However, these methods are not very suitable for color images and they change the shapes of object contours in large scales. In this paper, a novel density gradient-based method is proposed to solve these problems. The proposed method locates the edges using the density gradients of image pixels instead of the color/intensity gradients. Comparing with the traditional methods, the proposed approach can be applied more effectively to all types of images: gray level, color and multispectral images; and the detection results are invariant to the detection scales as well. It may find wide applications in computer vision and image processing.