Learning Appearance Primitives of Iris Images for Ethnic Classification

Iris pattern is commonly regarded as a kind of phenotypic feature without relation to genes. In our previous work, we argued that iris texture is race related, and its genetic information is illustrated in coarse scale texture features, rather than preserved in the minute local features of state-of-the-art iris recognition algorithms. In this paper, we propose a novel ethnic classification method based on learning appearance primitives of iris images. So we not only confirm that iris texture is race related, but also try to find out which kinds of iris visual primitives make iris images look different between Asian and non-Asian. In our scheme, we learned a small finite vocabulary of micro-structures, which are called iris-textons, to represent visual primitives of iris images. Then we use iris-texton histogram to capture the difference between iris textures. Finally iris images are grouped into two race categories, Asian and non-Asian, by support vector machine (SVM). Based on the proposed method, we get a higher correct classification rate (CCR) of 91.02% than our previous method on a database containing 2400 iris samples.