基于扩展二维不变矩的三维人脸特征提取

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

将二维不变矩理论和算法扩展应用于三维人脸特征提取的研究,在对三维人脸模型进行预处理及网格归一化的基础上,建立了具有平移、比例和旋转不变性的三维人脸模型的Zernike矩不变量特征表示,并基于距离可分性设计了一种特征评价测度,对所提特征和深度特征进行了对比评价。多姿态三维人脸数据库的实验结果表明,三维人脸模型的Zernike矩不变量特征具有对不同姿态人脸模型鲁棒、可分性强等优点。

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.