Using the cascade algorithm, the discrete approximation sequences of wavelet packet basis functions are computed. With the advantage of direct transform through correlation between wavelet packet basis function and the input image,the eigen-image correlation based recognition method is modified. The criterion for basis selection is changed for iris recognition. The multi-mother multi-vanishing moments joint best basis selection algorithm is proposed. After joint best basis selection,eigen-images are generated and applied in a volume holographic correlation recognition system to implement optical iris recognition. A good identification rate is obtained. For better recognition performance, optical filter of the best basis is designed and fabricated. The experimental result shows the optical wavelet packet gray scale filter promotes the recognition capacity of the system.