HUMAN FACE RECOGNITION METHOD BASED ON MULTI-LINEAR FUNCTIONS
Shuyue Ma · 2012
This paper undertakes the researches of face recognition problems in the conditions of different illuminations and viewing angles. To solve the problem of decreased human face recognition, this paper proposes a tensor-based multi-linear human face recognition method. First, we adopt the form of tensor to represent the image in human face database, and then, depending on different training individuals, we divide the original human face space into some subspaces, and obtain reflection transformation equations for their corresponding subspaces. In the recognition process, we can transform the reflection of the measured face images in different sub-spaces. Meanwhile, we have to calculate the amount of lost information before and after transformation. Thus, we have the human faces classified and identified. The main advantages of this method are reflected as follow: 1) Maintain the spatial relationships between the face images in the adjacent row vector and column vector by using tensor. 2) We can realize human faces recognition by adopting a simple linear transformation, with which we can have this algorithm been with a high recognition rate and a better computational complexity.