A quaternion gradient operator for color image edge detection

Abstract

Estimating gradients of color images is important to many color image processing tasks. However, the research on color image gradients is limited. In this paper, a novel method of estimating color image gradients is presented. The proposed gradient mechanism is based on measuring the squared local contrast variation of a color image function, in which the chromatic variation is evaluated by quaternion representation. As an application in color image processing, we apply the proposed gradient operator to the traditional grayscale image Canny operator for color edge detection. The edge detection results indicate that the proposed color gradient operator is superior to other state-of-the-art color image gradient methods.