Cascading Framework for Multi-level Image Quality Evaluation

Due to the light interference, mobile terminal iris recognition on the visible light cannot accurately extract texture information that used for iris recognition. In order to solve the problem, image quality evaluation is used to ensure that the iris image quality meets the recognition needs, find an excellent and complete iris structure, and extract the iris texture information to complete the matching. In this paper, a multi-level quality evaluation cascade framework is designed. This paper preprocesses the iris image, and then uses the Tenegrad evaluation function to complete the first level image quality evaluation, and then uses the wavelet transform and BP neural network to accomplish the second level image quality evaluation. To test the effectiveness of the algorithm, 900 iris images in the UBIRIS.v2 iris gallery are used for testing. The accuracy of quality evaluation is 96.1%. The result shows that the method can correctly evaluate the iris image quality, exclude the images that cannot be used for iris recognition, and effectively improve iris recognition on the mobile terminal.