Face Recognition of Illumination Tolerance in 2D Subspace Based on the Optimum Correlation Filter

Nowadays, face recognition technology has been greatly developed; for the problem of low recognition rate in the face recognition system of illumination variation, this thesis proposes the face recognition method of illumination tolerance in two-dimensional subspace based on the optimal correlation filter. Firstly, through the use of a particular class 2D-PCA the face image is reconstructed and by using the optimum projecting image correlation filter (OPICF) the faces images will be tested to project into the two-dimensional subspace; secondly, refactoring correlation filter (RCF) is used to refactoring the image and the face images are classified according to pre-set threshold of tolerance illumination; finally, the performance of the proposed method is evaluated on two major face databases like Yale B and PIE. Experimental results show that compared to other existing correlation filters, filter system designed in this paper has the inherent tolerance of the illumination changes and emphasizes the spatial high frequency components of refactoring face images consisting of facial features. Meanwhile, the refactoring face images don’t consider the illumination variations, so the proposed method reduces the low-frequency component in the spaces and strengthens the recognition accuracy rate under the situation of illumination variations. Compared to the performances of other filters, it has a constant high accuracy and low error acceptance rate.