3D face feature extraction based on extended 2D invariant moment

A novel 3D face feature extraction method based on 2D extended strategy is proposed. Zernike Moment Invariants of multi-pose 3D face model, which have scale, translation and rotational invariance, are extracted after preprocessing and grid normalization of 3D face models. A suitable feature evaluation measure is then proposed based on the parameter space and is applied to the evaluation of Zernike Moment Invariants and depth feature. Experimental results based on multi-pose 3D face database show that the proposed method is robust to pose variant and has a separability.