基于（2D）^2NMF及其改进算法的人脸识别

Face recognition based on （2D）^2NMF and its improvement

非负矩阵分解(NMF)是基于部分的特征提取方法,能够克服局部遮挡和光照问题,在图像识别任务中效果较好。然而传统算法中,NMF提取的特征是非正交的,且二维图像常被向量化处理,不仅丢失一些结构信息,还导致了数据的高维,不利于提高识别精度和速度。利用图像矩阵取代传统的图像向量表示,提出新的(2D)2NMF方法提取二维图像特征,并通过特征正交化和图像变形等措施,改善了算法性能。人脸识别实验表明,上述措施能够有效提高识别的精度和速度。

Non-negative matrix factorization （NMF） is an effective method for parts-based feature extraction, and it ean deal with partial occlusion and some illumination problems. However, the bases learned via NMF are not orthogonal. Translating an image into a vector often loses the structure information of pixels and leads to a high dimensionality. By adopting the image matrix instead of image vector, （2D）2 NMF method was presented to extract the 2D features of an image, so the performance of the method was improved by the NMF bases orthogonalization and image diagonalization. The experimental results show the improved performance compared with the traditional NMF and 2DNMF.