Abstract: Conventional edge detection algorithms suffer from the problems of color edge deletion, low positioning accuracy, a false pseudo-edge and poor anti-noise performance. This paper proposes an improved method based on quaternion Canny operators to solve such problems. The method utilizes colored edge features quaternion matrix to calculate image edge gradient amplitude and thus can extract the edges accurately. We use PSNR and Pratt quantitative to evaluate the quality of the proposed algorithm, and the experimental results demonstrate that it is insensitive to noise and can reduce the color edge Information missing rate and false detection rate efficiently.