Iris recognition for partially occluded images: Methodology and sensitivity analysis

Accurate iris detection is a crucial part of an iris recognitionsystem. One of the main issues in iris segmentation is coping withocclusion that happens due to eyelids and eyelashes. In theliterature, some various methods have been suggested to solve theocclusion problem. In this paper, two different segmentations ofiris are presented. In the first algorithm, a circle is locatedaround the pupil with an appropriate diameter. The iris areaencircled by the circular boundary is used for recognitionpurposes then. In the second method, again a circle is locatedaround the pupil with a larger diameter. This time, however, onlythe lower part of the encircled iris area is utilized forindividual recognition. Wavelet-based texture features are used inthe process. Hamming and harmonic mean distance classifiers areexploited as a mixed classifier in suggested algorithm. It isobserved that relying on a smaller but more reliable part of theiris, though reducing the net amount of information, improves theoverall performance. Experimental results on CASIA database showthat our method has a promising performance with an accuracy of99.31%. The sensitivity of the proposed method is analyzedversus contrast, illumination, and noise as well, where lowersensitivity to all factors is observed when the lower half of the irisis used for recognition.