IRIS RECOGNITION METHODS | Thesis Guidance in Chandigarh

IRIS RECOGNITION METHODS

Daugman method

The method of Daugman uses an integral-differential operator designed for locate the circular iris and pupil regions with the upper and lower eyelids arcs. The operator finds the circular path in which there is more variation in pixel values with the radius and centre x-axis and y-axis position in the circular contour. The operator is functional iterative by the smoothing progressively amount lessen for attaining the exact localisation. The eyelids are localized in the same manner, by contour integration path that is changed from circular to an arc.

Active Contour Models

Ritter et al. has used active contour systems for pupil localization in eye images. The active contours answer to fixed external and internal forces for moving across an image or deforming internally till equilibrium state is achieved. The contour has few vertices that have positions that can be changed by an internal force, two disparate forces, and dependent on the preferred. For the pupil region localization, the internal forces are being calibrated for forming a contour globally growing discrete circle. The external forces are basically uses edge information. For improving the accuracy, Ritter et al. has used the variance image as compare to the edge image. A point inside the pupil is located by means of variance image and later, a discrete circular active contour (DCAC) is being established by this point as its centre. The DCAC is moved in the internal and external influence forces till it achieve equilibrium, and then the pupil is being localized.

Eyelash and Noise Detection

Authors, Kong and Zhang, has presented a technique for detection of eyelash, from where the eyelashes are taken as two types connection which has separate eyelashes and are isolated in an image with multiple eyelashes. These are bunched jointly with an overlap of an eye image. The separable eyelashes are being found out by 1D Gabor filters as the convolution of different eyelash with the Gaussian smoothing function being resulted in low output value.So, if a resultant point is same as a threshold, then it is clear that the point belong to an eyelash. Different eyelashes are being detected by the variance of intensity. So, if the intensity values variance in a little window is less than threshold, then the centre of the window is being measured as a point in an eyelash. Authors, namely, Kong and Zhang model has also utilized connective criterion, for each point in the eyelash that should be connected to other point in the eyelash/eyelid.  The specula reflections within the eye image are being detected by using thresholding, as the intensity values destined at the regions are higher than at another regions presented in the image.

Daugman’s Rubber Sheet Model

Rubber sheet model is devised by Daugman that remaps every point in the iris region in the polar coordinates pair (r,θ) in which r is on the interval [0,1] and θ is now the angle [0,2π]. These rubber sheet systems convert Cartesian coordinates in the form of polar coordinates. The Cartesian coordinates communicate to pupil coordinates with the iris boundaries next to the θ direction. The rubber sheet model has account pupil dilation with the size inconsistencies for producing normalized representation within the constant dimensions. The iris region is modeled as a flexible rubber sheet anchor at the boundary of iris with the pupil centre as anorientation point

Virtual Circles

In terms of Boles system, the iris images are initially scaled for having constant diameter for the comparison of two images first is known as the reference image. It operated differently for the methods, as the normalization cannot be performed till the matching is for two iris regions, than the performance of normalization and the results are saved for next comparisons. The two irises with the similar dimensions, the features are being extracted from the iris region with the storage of the intensity values with the virtual concentric circles, by the centre origin of the pupil. The normalization resolution is being selected; then, the variety of data points are being extracted from each iris is considered same. This is basically the similar as Daugman‟s rubber sheet technique, however, the scaling is lying at match time, and is considered as relative for the comparison of iris region, than the scaling for several invariable dimensions.

Haar Wavelet

Author, Lim et al. has used wavelet transform for extracting the features from the iris region. As of multi-dimensional filtering, the feature vectors with 87 dimensions are computed. Because every dimension have the real value starting from -1.0 to +1.0, the feature vector is considered as sign quantized, thus, the positive value is shown by 1 with the negative value as 0. The outcomes being in a compact biometric template consists of only 87 bits.

Summary
Review Date
Author Rating
51star1star1star1star1star

Leave a Reply

Your email address will not be published. Required fields are marked *