Quaternion moment and its invariants for color object classification

Moment invariants are important shape descriptors in computer vision. In this study, we propose new sets of quaternion moment descriptors for color image. They are constructed in the quaternion framework and are an extension of complex moment invariants for grayscale images. This is a useful tool in color image processing and color object recognition tasks that require the similarity invariance. The advantage of the proposed quaternion moment invariants is that they can not only process color image in a holistic manner but also grayscale one. In addition, the computational complexity of the proposed method is much lower than the quaternion Zernike moments defined in the polar coordinates. An example of using the quaternion moment invariants as pattern features for a color object classification application is given. Theoretical and experimental results show that the proposed descriptors perform better than the other competing moment-based methods.