Face Recognition Using Cubic B-Spline Wavelet Transform

Face recognition using cubic B-spline wavelet transform is proposed in this paper. The proposed scheme is based on the analysis of face recognition using wavelet transform and the viewpoint that detail subbands after wavelet transform also have a lot of feature information. Then, the feasibility of a new application that cubic B-spline wavelet use Mallat algorithm to decompose an image under some special circumstances is also presented in this paper. The concrete algorithm can be roughly listed as follows: At first, cubic B-spline wavelet is used to decompose each face image at suitable levels to produce an approximation subband and three detail subbands at the last level decomposition. Then Fourier transform and PCA are performed on several optimal subbands selected from four subbands in succession, the last result will be produced by weighted multi-distances fusion. The proposed algorithm is tested on face images that differ in expression, illumination or pose separately, obtained from JAFFE, Yale and UMIST face databases. It is surprising finding that the proposed algorithm has significant performance, especially under the condition of illumination perturbations.