摘要
人脸图像的校正是人脸识别相关技术的前提性工作之一。眼镜作为一种常见的干扰物对识别结果有着重要的影响。针对该问题,提出了一种去除眼镜的面部图像校正方法。首先对无眼镜的一组人脸样本集进行训练,然后通过主成分分析法对输入的带眼镜的人脸图像进行重构,再按照原始图像的灰度特征对重构出的图像进行处理,最后合成为不带眼镜的人脸图像。实验结果表明,该算法易于实现,合成效果比较自然,有利于后续的处理过程。
The rectification of facial images is one of prerequisite research works in face recognition. As a common factor, the eyeglasses interference has an important impact to the effect. Aiming at eyeglasses removal in face images, a rectification method is proposed: Based on facial images training samples, input facial image without eyeglass is reconstructed by principle component analysis, then the reconstructed image is processed according gray features of original image, finally, facial image without eyeglasses is synthesized. Experimental results show that the subtle synthesized effect is easy to be realized, and the removal of eyeglasses in face images will be propitious to post processing.
出处
《计算机工程与设计》
CSCD
北大核心
2009年第10期2480-2481,2589,共3页
Computer Engineering and Design
基金
陕西省教育厅自然科学研究基金项目(08JK290)
陕西省自然科学基金项目(2006F34)
陕西省自然科学基金研究项目(SJ08F32)
关键词
人脸图像
合成
灰度
眼镜
校正
face images
synthesis
gray
eyeglasses
rectification