Abstract: A novel approach to iris image sequence quality assessment is proposed. Firstly, two assessment image blocks of each input iris image are extracted by using image projection and binarization method. Secondly, values of a two-dimension complexity measurement called 2D-C⁰ complexity and Fourier frequency spectrum of the assessment image blocks are computed as features. Finally, a classifier uses these features to select the clearest iris image for personal identification. Experiments were made by using K-NN, BP neural network and support vector machine classifiers based on these features, and the experimental results are satisfactory.