摘要

A detection method for motion-blurred iris images was developed based on ellipse fitting to make the iris recognition system more accurate. The positions of luminous spots were estimated preliminarily in iris image by Gabor filter, then the actual luminous spot was detected in iris image which has been magnified to 5 times. With the ellipse fitting done to the boundary of every luminous spots, the fitting error and the ratio of major axis to minor axis are both calculated. An image could be judged as a motion-blurred one if the ratio is greater than 1.55 with relatively small fitting error. Results of the experiments carried out on CASIA-IrisV3-Lamp database indicated that the accuracy of detection of motion-blurred images is up to 100%, thus improving the accuracy of the whole iris recognition system.

全文