Abstract: Iris location, an important link in iris recognition, is to detect inner and outer boundaries of iris in an iris image, In the paper, an improved algorithm is proposed to locate iris by means of curve fitting. In this algorithm, an iris image is blocked firstly and its sub-image blocks which cover pupil are extracted. Next, a binary threshold of pupil is set by means of the method of maximum variance between clusters, and the inner-boundary-points of iris are determined. And then, the outer-boundary-points of iris are gotten on the basis of gray gradient of key regions of interest. Consequently, some characteristic pixels in regions of interest are selected respectively and inner and outer boundaries of iris are fitted according to principles of least squares. The results of experiment show that the proposed algorithm can locate iris accurately and quickly.