Iris Recognition Based on Wavelet Transform with Shift Invariance Preprocessing

The normal discrete wavelet transform is sensitive to the initial position of asignal and can not be used to extract stable iris features. To weaken the negative effect on the recognition result due to the shift sensitivity of the wavelettransform，a shift invariance preprocessing method based on directional energy distribution sequences is proposed to correct the angular rotation of different iris textures, which corresponds to the horizontal shift of normalized iris blocks. A threshold scheme is then performed to transform all the wavelet coefficients into a binary vector to represent iris features. In verification mode, the weighted Hamming distance classifier is finally adopted to perform multi-template matching to identify the unknown feature vector. Comparison experiments show that the proposed algorithm enhances the availability of iris feature representation by wavelet transform and leads to a good performance of iris recognition.