单张人脸图像眼镜边框摘除方法

Method for Eyeglasses Frame Removal from Single Facial Image

眼镜对人脸图像的特征提取和识别有重要的影响.本文提出了一个基于单张人脸图像的眼镜边框摘除方法.首先利用自适应双阈值Canny算法提取眼镜边框的准确边缘,以此确定待修复区域并构造修复模板,然后利用整体变分(TV)算法进行目标修复,从而得到没有眼镜边框的人脸图像.实验结果表明,本方法能够成功摘除眼镜边框,效果良好.

Eyeglasses have important impact on face image feature extraction and recognition. A method is proposed for eyeglasses frame removal from single facial image in this paper. Firstly, it used the adaptive dual-threshold canny algorithm extract accurate edge of eyeglasses frame to determine the area to be repaired and construct repair template. Then the total variation（TV） algorithm is used to repair target area and obtain facial image without eyeglasses frame. Experimental results show that the proposed method can be successfully removed the eyeglasses frame with good effect.