Face Recognition Using Two-Dimensional Tunable-Q Wavelet Transform

Abstract

Tunable-Q wavelet transform (TQWT) is a discrete wavelet transform that has been very effective in decomposing signals with oscillatory nature. In this paper, we develop a new two dimensional tunable-Q wavelet transform (2D-TQWT) using its 1D prototype and propose an approach for face recognition using 2D-TQWT. The proposed approach decomposes a face image into four sub bands. This is followed by extraction of local binary pattern based histogram features from different sub-bands. This extracted information is further combined to get the final representation. In order to evaluate the performance of the proposed 2D-TQWT based face recognition approach, experiments are carried out on two datasets namely, Yale and ORL face datasets. The performance of proposed approach is also compared with other existing wavelets. Experimental results show that the 2D-TQWT yields better recognition accuracy than other wavelets employed in our experiments for comparison.