结合形状滤波和几何图像的3D人脸识别算法

Three dimensions face recognition by using shape filtering and geometry image

表情变化是3维人脸精确识别面临的主要问题,为此提出一种新的对表情鲁棒的匹配方法。通过形状滤波器将人脸空域形状分成不同频率的3个部分:低频部分对应表情变化;高频部分代表白噪声;包含身份区分度最大的中频信息作为表情不变特征。再利用网格平面参数化,将人脸网格映射到边界为正四边形的平面区域内,经过线性插值采样得到3维形状的2维几何图像。最后通过图像匹配识别人脸。FRGC v2人脸数据库上的实验结果表明,使用形状滤波能显著提高算法的精度和鲁棒性。

Achieving high fidelity in the presence of expression variation remains one of the most challenging aspects of 3D face recognition. In this paper, we propose a novel recognition approach for robust and efficient matching. The framework is based on shape processing filters that divide face into three components according to its frequency spectra. Low-frequency band mainly corresponds to expression changes. High-frequency band represents noise. Mid-frequency band is selected for expression-invariant feature that contains most of the discriminative personal-specific deformation information. After bijectively mapping facial mesh into square domain based on mesh parametrization, we obtain 2D geometry image of 3D shape with linear interpolation for face matching. We conduct extensive experiments on FRGC v2 databases to verify the efficacy of the proposed algorithm, and validate that by using shape filter, it offers a performance improvement for both accuracy and robustness.