Reduced robust facial feature descriptor using DTCWT and PCA

This paper present a robust reduced facial feature descriptor for face recognition by using dual tree complex wavelet transform and principal component analysis. Proposed approach uses extra dyadic down sampling strategy on coefficient of DT-CWT to reduce the size of feature vector and further without loss of generality principal component analysis is used on reduced feature vector significantly. Geometrical structure in facial image can be represented efficiently and effectively with low redundancy by using extra dyadic down sampling strategy. To extract facial feature this method is robust against the discrepancy of shift and illumination than the DWT. It has been verified experimentally that the proposed method is more dominant to reduce the size of feature vector.