人脸识别中局部遮挡处理技术研究

Study of Partial Occlusion Treatment Technology in Face Recogniton

局部遮挡严重影响了人脸识别准确率,提出了自动最优局部遮挡的重建模型.该模型首先利用人脸图像的灰度值判断人脸遮挡类型,然后利用M估计器(M-estimator)对遮挡掩模进行估计,找到最佳局部遮挡的人脸重建系数,重建人脸,减弱人脸遮挡的影响.实验证明了局部遮挡处理的有效性,明显提高了识别准确率.

Partial occlusion has seriously affected to the accuracy of face recognition.A model about automatically partial occlusion reconstruction is proposed.First,the face gray value is used to determine the partial occlusion type,and then a M-estimator is used to estimate the occlusion mask,the best face reconstruction coefficient is found to reconstruction occlusion area and weaken the impact of partial occlusion.The experiments have proved the effectiveness of the partial occlusion treatment,and the significantly improvement of the recognition accuracy.