一种基于遮挡区域补偿的眼镜去除方法  

Eyeglasses removal method based on occlusion region compensation

作　　者：胡国靖 娄震[1] 

机构地区：[1]南京理工大学计算机科学与工程学院,南京210094

出　　处：《计算机应用研究》2013年第12期3863-3865,共3页Application Research of Computers

摘　　要：为了提高戴眼镜人脸图像的识别率,提出了一种从人脸图像中检测并去除眼镜的方法。首先对输入的戴眼镜人脸图像与系统预留的无眼镜人脸图像进行基于人眼位置的标定,检测出眼镜遮挡区域,再用无眼镜人脸图像中对应的遮挡区域对戴眼镜人脸图像进行补偿,从而合成了对应输入图像的不戴眼镜的人脸图像。实验结果表明,该方法能有效地合成无眼镜人脸图像,将合成后的人脸图像再应用于人脸识别系统,识别率显著提高。This paper proposed an improving method to detect and then removed eyeglasses from facial image. First, it obtained the occluded region by calibrating the facial image with eyeglasses and the reserved one without eyeglasses. After the occlusion detection, it synthesized the facial image without eyeglasses by compensating the facial image which had eyeglasses with the occluded region. The experimental results show that this method can effectively synthesize facial image without eyeglasses and the recognition rate improves significantly when the synthesized image is applied to face recognition system.

关 键 词：人脸识别 眼镜去除 图像标定 遮挡区域检测 图像合成 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]