采用图像融合技术的多模式人脸识别

Multimodal Face Recognition Based on Images Fusion

利用图像融合技术实现了基于可见光图像和红外热图像相结合的多模式人脸识别,研究了两种图像在像素级和特征级的融合方法。在像素级,提出了基于小波分解的图像融合方法,实现了两种图像的有效融合。在特征级,采用分别提取两种识别方法中具有较好分类效果的前50%的特征进行特征级的融合。实验表明,经像素级和特征级融合后,识别准确率都较单一图像有很大程度的提高,并且特征级的融合效果明显优于像素级的融合。因此,基于图像融合技术的多模式人脸识别,有效的增加了图像的信息量,是提高人脸识别准确率的有效途径之一。

A multimodal face recognition technology based on visual and infrared images fusion is introduced, and fusion methods on pixel level and feature level are discussed. On pixel level, image fusion based on wavelet decomposition is used to effectively combine pixels of visual and infrared images to form new images. On feature level, a new feature extraction method is proposed by ranking and extracting top 50% of feature vectors. The recognition experiment shows that correct recognition rate （CRR） is improved by using pixel fusion and feature fusion, compared with that using single type images. Further more, the feature fusion of images is obviously superior to the pixel fusion. In conclusion, multimodal face recognition based on image fusion combines information from images of two types, and become to effective method for increasing CRR.