The Research On Iris Image Quality Assessment

The accuracy of the iris recognition system is closely related to the imagequality, low quality images will seriously affect the whole system’s accuracy. Thepurpose of iris image quality assessment is to filter out this kind of images, andguarantee that the quality of the images for subsequent operation, and then improvethe system recognition accuracy. The study of iris image quality assessment startedlater, although it has achieved some results, there is still no assessment system havewon the approval of academic circles, therefore, the study on iris recognitionassessment is very necessary.The main work of this paper includes the flowing aspects:(1)Through analyzing the various anomalies, the iris image anomalies wereclassified into three categories in this paper: incomplete iris area, low iris visibility,low iris definition. Aimed on the above three situations, this paper used iris integrity,iris visibility and iris clarity as indexes to do image quality assessment.(2)Through the comparison and the analysis of the existing iris image qualityevaluation technology, this paper proposed an iris image quality assessment methodbased on gradation feature, and detected the iris integrity with pupil position,detected the iris visibility with eyelid occlusion and pupil zoom ratio, detected theiris clarity with the average height of pupil edge.(3)The ROI-based image quality assessment method was used in iris imagequality assessment, use the pupil area as the iris integrity detection ROI, use therectangular area above the pupil as the iris visibility detection ROI, use the ringbetween the pupil and iris as iris clarity detection ROI. Get the iris image qualityevaluation index through analyzing the ROI.(4)The proposed methods and Daugman method was tested with VC on 100image samples, the results show that the proposed methods are better than Daugmanmethod both in speed and accuracy, and the speed of gradation feature-based method is faster than ROI-based method, the accuracy of the gradation feature-based methodis lower than ROI-based method.