Face blind separation using wavelet packet independent component analysis

A novel wavelet packet based approach to Subband decomposition independent component analysis (SDICA) is proposed. The mutual information based on small cumulant is introduced to select the Subband with least dependent components. We present favorable comparisons to the WPSD ICA and other ICA algorithm in extensive simulations. We demonstrate consistent performance in terms of accuracy and robustness as well as computational efficiency of WPSD ICA algorithm. Experimental results demonstrate that the proposed method can significantly improve the face recognition performance.