FACE RECOGNITION BY PARALLEL PARTIAL PROJECTIONS

摘要

A novel approach is proposed for a robust face recognition using eigenfaces. It is known that robust algorithms are necessary for the face recognition in natural environments and one feasible scheme can be constructed by the robust (estimation of the) projection to the eigenfaces. However, the robust-projection methods often suffer from the breakdown point problem, and the recognition scheme cannot tolerate excessive noise below the point. In order to solve the problem, we propose a novel recognition scheme without using a robust projection. Our scheme uses a set of simple partial projections instead of the robust projection, which is followed by a robust discrimination. In order to compare the proposed scheme with the conventional scheme, a set of experiments are performed on the Yale Face Database B. Experimental results show that the proposed scheme works better even when the eigenfaces are constructed from somewhat noisy images and input images involve a considerable amount of noise. The proposed scheme also gains an advantage of computational cost over the conventional scheme.