| 摘要: |
| 研究了基于奇异值分解的人脸识别问题。首先证明了图像的大量信息主要体现在图像矩阵奇异值分解的前几个最大奇异值所对应的左、右奇异向量中,然后给出了模板图像的基图像,并将图像展开成基图像的线性表示,提取其组合系数作为图像的代数特征并用于人脸识别中。实验表明,较以投影系数向量为代数特征的人脸识别方法,该方法所需的运行时间明显降低,而且与基于奇异值向量作为图像特征的方法相比,该方法的识别精度明显提高。 |
| 关键词: 计算机视觉,模式识别,人脸识别,奇异值分解,基图像 |
| DOI：10.3969/j.issn.1001-893X. |
| Received:July 26, 2007Revised:November 26, 2007 |
| 基金项目:重庆市教育委员会科学技术研究资助项目(KJ070511)，广东省自然科学基金资助项目(D06300640) |
|
| Face Recognition Based on the Basic Images of Singular Value Decomposition (SVD) |
|
| () |
| Abstract: |
| This paper researches the question of face recognition based on singular value decomposition(SVD).It is firstly proved that most important information of image is contained in the left and right singular vectors corresponding to some largest singular values.Then the basic images of template image are presented and an image is expressed with the linear combination of the basic images.The combined coefficients are extracted as algebraic feature of image and are applied to face recognition.The experiment results show that the run time is remarkably reduced compared to the method based on the projection coefficient vector,and the recognition rate is remarkably increased compared to the method based on the singular feature vector. |
| Key words: computer vision,pattern recognition,face recognition,SVD,basic image |
