眼镜遮挡下的正面人脸识别

针对眼镜遮档对人脸识别影响较大这一问题，提出一种从正面人脸图像中提取并摘除眼镜的方法。首先利用主成分分析和独立成分分析法对输入的戴眼镜人脸进行重建，对比重建人脸和输入人脸，从而提取眼镜遮档区域;然后经过迭代误差补偿合成相应的无眼镜人脸;最后考虑到合成图像的特殊性，使用改进的特征加权方法实现人脸识别。实验结果表明，利用提出的人脸重建和特征加权方法进行戴眼镜人脸识别，正确率可以达到91%，优于传统方法。

Frontal Face Recognition under Glasses Occlusion

Aiming at the problem that eyeglasses occlusion severely affecteds the recognition rate, this paper described a method to remove eyeglasses from the frontal facial image. Firstly, the input facial image was reconstructed by PCA+ICA and then the region occluded by the eyeglasses was obtained by comparing the reconstructed facial image and the input facial image. Secondly, an eyeglasses facial image was synthesized by recursive error compensation. Finally, considering to the specialty of synthetic image, the improved feature weighted method was used to realize the face recognition. hhe experimental results show the method is simple and easy to realize, can generate natural looking facial images without eyeglasses. The average accuracy of face recognizing is 91%, and this method outperforms traditional methods.