Iris Recognition Based on the Barycenter Distance Vector of New Non-Separable Wavelet

This paper makes an attempt to analyze the local feature structure of iris texture information based on the barycenter distance of new non-separable wavelet. When preprocessed, the annular iris is normalized into a rectangular block. Several non-separable wavelet filters are used to capture the iris texture. In every filtered subband coefficients, we extract a certain number of largest positive coefficients and smallest negative coefficients that can represent the local texture most effectively in each subband. The barycenter of these positive coefficients in each subband is called positive barycenter, and the barycenter of negative coefficients is called negative barycenter. Then, the vector from negative barycenter to positive one is called barycenter distance vector, which is regarded as the iris feature vector. Iris feature matching is based on the similarity of the vectors. Experimental results on public databases show that the performance of the proposed method is as good as Daugman´s method, and our method is more robust than Daugman´s method to rotation transform in small scale.