基于三维重建的人脸姿态估计

人脸姿态估计是计算机视觉及人脸识别领域的一项关键研究内容.将三维模型应用于估计人脸图像姿态参数,提出了基于三维重建的人脸姿态估计算法.根据人脸正面图片上的特征点计算出形状参数,实现三维人脸重建.基于三维人脸模型,由姿态图片上提取的特征点信息推知等姿态下模型上对应特征点的信息,针对照模型正面姿态,运用线性回归估计姿态参数.实验表明,重建的三维人脸具有较好的真实感,在较大的姿态变化范围内,该方法也能够取得较好的估计精确并具有鲁棒性.

Face pose estimation based on 3D face reconstruction

Face pose estimation is a key research topic in the fields of computer vision and face recognition. Applying the 3D face model to estimate the pose of a 2D image, an approach based on 3D face reconstruction was proposed. Firstly, the shape parameter was calculated based on the feature points of a frontal face image, thus realizing the 3D face reconstruction. Then, based on face model, the information of feature points in the 3D model was deduced by the corresponding points in the input image. And by contrast with the 3D model of frontal pose, the liner regression was used to estimate the face pose. Experiments show that the reconstructed 3D model is realistic, and the estimation method achieves better accuracy and robustness in a wide range of head motion.