Iris recognition using localized Zernike moments

摘要

A system receives an iris image and segments the iris region. The segmented iris region is mapped to a unit disk and partitioned into local iris regions (or sectors) as a function of the radius and angle The system calculates localized Zernike moments for a plurality of regions of the unit disk. The localized Zernike moment includes a projection of the local iris region into a space of Zernike polynomial orthogonal basis functions. The system generates an iris feature set from the localized Zernike moments for each partitioned region, excluding the regions which are comprised by occlusion. The iris features are weighted based on the conditions of blur, gaze and occlusion of the iris region. A probe iris image is then matched to a plurality of iris images in a database based on the distance of its feature set to the corresponding plurality of iris feature sets.

主权项

1. A system comprising:
a computer processor and computer storage device configured to:
receive an image of an iris;segment the image of the iris, thereby creating a local iris region;map the image of the iris as a unit disk starting at an inner boundary of the iris and stopping at an outer boundary of the iris;segment the unit disk as a function of a plurality of Gaussians centered at different locations in the local iris region, thereby generating a plurality of iris sub-regions;calculate a localized Zernike moment for each of the plurality of iris sub-regions to obtain a plurality of localized Zernike moments, each localized Zernike moment comprising a projection of the local iris region into a space of Zernike polynomial orthogonal basis functions;select the localized Zernike moments calculated for the iris sub-regions such that the iris sub-regions are not compromised by occlusion;generate a local iris texture feature set using the selected localized Zernike moments calculated for one or more iris sub-regions;classify the image of the iris as a function of the local iris texture feature set; andcompare the classified image of the iris to a plurality of iris images in a database.