A Shape Descriptor Using Complex Moment Invariants

This paper proposes a robust and effective shape feature, which is based on a set of orthogonal complex moments of images known as Zernike moments. Zernike moment phase is usually not used in image description since it’s sensitive to image rotations. However, phase captures important image information, which is revealed by our numerical analysis of image reconstruction. We therefore propose combining both the magnitude and phase coefficients to form a new shape descriptor, named as invariant ZM descriptor (IZMD). Its scale and translation invariance of IZMD are obtained by pre-normalizing the image using the geometrical moments. To make the phase invariant to rotation, we propose a method of phase correction. Experiment results show that IZMD is, in general, robust to changes such as rotation, translation and scaling. It also outperforms the commonly used magnitude-only Zernike moment descriptor in noise robustness and object discriminability. Keyword: Zernike moment, invariants, shape matching