텐서 기반 다선형 주성분 분석기법에 의한 얼굴 인식

Abstract

This paper is deals with face recognition and representation using Tensor-based Multilinear Principal Component Analysis (T-MPCA) approach. This method extracts feature directly from the tensor representation rather than the vector representation. Many images and video data are naturally tensor objects. For example, colour images are 3D (third-order tensor) objects with column, row, and colour modes. Currently, most face recognition algorithms are performed by 2D gray images. Therefore, video sequences with gray image can be viewed as third-order tensors with column, row, and time modes. In order to evaluate the performance, we perform face recognition based on face database constructed in test bed and office environments. The experimental results on face database reveal that the presented method shows a better performance in comparison with the well-known methods in distant-varying environments.