图像集的离散度矩阵及其在人脸图像识别中的应用

本文提出了图像集离散度矩阵的基本概念及用于人脸图像识别的方法。该方法把同一类图像集中的各幅图像的信息矩阵的奇异值向量作为矩阵中的一列构成图像集特征矩阵。把这些矩阵组成一图像库，并找出图像集离散度矩阵和图像库离散度矩阵，从而计算出它们的最佳投影系数。然后，把测试样本的图像集特征矩阵与图像集库中的训练样本图像集的特征按投影系数相比较找出它的相似程度，从而进行人脸图像识别。

FACE RECOGNITION BASED ON SIMILARITY OF IMAGE SETS

This paper puts forward a new approach of face - image recognition based on the scatter of image sets in accordance with the cognitive law. This approach considers the singular vectors of information matrix of each image in the image set as columns of the characteristic matrix of the image set, and composes image - sets base of these matrices. Scatter matrix of image set and scatter matrix of image - sets base will be calculated. Then, it compares the characteristic matrix of the testing samples with that of the training samples in the image - sets base by projective coefficient, to find out the degree of their resemblance - the similarity of image - sets. In this way, the face recognition can be carried out.