In order to eliminate the flaws in traditional iris detection methods, a new optimization algorithm is proposed, which is based on the theory of the dynamic contour model and used to revise the raw contours computed by Hough circle detection. First, we obtain a raw estimate of the iris contour by Hough algorithm. And then based on the idea of calculus, an iris image is segmented into a number of small rectangles which have the same size. After that we establish a dynamic contour model containing smooth function and gradient function in order to revise raw contours in each rectangle and achieve a precise estimate. Finally, we use the proposed algorithm, Hough algorithm and kmeans clustering to achieve respective contour estimate in experiment. The result shows that the proposed method have a large advantage over the Hough circle algorithm in terms of precision, and it also have a higher computational efficiency than kmeans especially in dealing with images which size is over 300×200.