In the past two decades, quaternions have been more and more used in color image processing to represent color images by encoding three channels into the imaginary parts of quaternion numbers. The main advantage of this quaternion representation is that a color image can be treated holistically as a vector field. We have also used it to color image denoising and object recognition. I will present two of our works: one is a new quaternion filter for removal of Gaussian noise in color images; the other one is quaternion Zernike moments and their invariants for color image analysis and object recognition.