Iris Recognition for Personal Identification Using Gabor Filters

In this work, an identifying and recognizing person technique is presented based on its eye’s iris. The introduce technique utilizes several image processing methods; e.g. edge detection, contour following, chain coding, Gabor filters, image normalization, feature extraction, etc. JPEG images of size 256×256 pixels are converted into grayscale form by averaging their Red, Green and Blue bands.The produced grayscale image then traced into edges and boundaries using Marr-Hildreth edge technique with an appropriate Gaussian standard deviation value (i.e. σ =3). Iris region is detected by performing a suggested contour following method which is based on the chain coding algorithm.Eight Gabor’s filtering channels are simulated (with different orientations;ooooooo05.157   and ,135,5.112,90,5.67,45,5.22,0) and convolved with the iris’s extracted image. The feature’s globalized images produced from the Gabor filters convolution then, normalized to suggested mean and variance values, using a suggested normalization method.  
A total of 40 features; (eight globalized images, and five for the sliced sectors), are stored in a created database file.Person identification can then carried out, for new eye’s iris, by following the same procedures (mentioned above), but comparing his or her extracted 40 features with those already preserved in the database file. The test of identification results in either “Identified Person” or “Undefined Person” depends on the similarity matching operation which carried out by utilizing the Minimum-Distance Classifier “MDC” criterion adopted in this research.The result of using Gabor filters alone has been found as to be copying better than utilizing the gradient Sobel filter with Gabor filters; i.e. The Gabor results alone yield higher variance values between different persons. The classification accuracy = 100% using Gabor filters alone, while the classification accuracy =80% using Sobel filter with Gabor filters.