1. UKM Learning and Research Repository
2. Master Thesis / Tesis Master
3. Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Title:	Iris video segmentation based on dynamic programming
Authors:	Nur Khder Nseaf (P72222)
Supervisor:	Azrulhizam Shapi'i, Dr.
Keywords:	Video compression Optical pattern recognition Universiti Kebangsaan Malaysia -- Dissertations Dissertations, Academic -- Malaysia
Issue Date:	27-Dec-2016
Description:	Biometrics system as used in the identification, at the present time the iris recognition based on video is growing rapidly, has seen iris recognition to be of value for its reliability known and precision that plays a big role iris recognition accuracy in determining the success in determining an individual's identity. However, the most challenging obstacles that hamper the accuracy of iris recognition are separation the iris from the eyelids and eyelashes and the process of detecting the boundaries of the iris. In order to obtain a robust identification system, it requires a fast and accurate iris segmentation method to improve the recognition rate. Based on the aforementioned problem, a study was carried out To propose a faster and accurate pupil segmented based on dynamic threshold and segment both of IRIS and eyelid simultaneously based on dynamic programming for non-ideal iris video scanner. In this study, this approach was utilized in the segmentation stage (i.e. location of iris region in eye image). The performance of this approach emphasizes on the precision of iris segmentation on the basis of determining the pupil boundary by using dynamic threshold with morphology operation and Hough transform to determine the center and radius of pupil to use this center as initial center for outer Iris and use initial radius for outer iris that coverage the Iris region from experimental analysis the initial radius is equal 150, followed by normalized the initial outer boundary and then apply canny edge detection to initial normalized outer boundary, the key here is The strongly connected edges between pixels in polar domain are detected using optimal minimum cost path dynamic programming (DP) technique that is likely to be the Iris and eyelid boundary. Followed this eyelash removal to extract iris free noise. Following this is normalization of the iris region that’s extracted into an oblong block having continuous dimensions. This stage aimed at accounting the imaging inconsistencies by applying Daugman’s rubber sheet model. The process of extracting and encoding features aiming at creating a template distinguishing iris by applying Log-Gabor Filters. Finally, the process of matching was conducted using Hamming Distance Test in the matching stage. The study also used the database of 2230 iris images of format .jpg created from MBGC v1 NIR Iris Video database this database are NIR 290 videos with 113 subjects and with 640x480 pixel frame dimensions. Each video consists of an eye either the left or right only. The image of the eye represented the input to the database, while the produced iris template represented the output produced that mathematically represents the identified iris region. The result of the evaluation showed that the performance of accuracy for our proposed optimal minimum cost path dynamic programming technique is 99.6637% with time execution is 0.8241875 second, while accuracy for Hough Transform based (masek 2003) method is 95.9459% with 18.135305 second of time execution, and result for Geodesic Active Contour Algorithm by (Shah& Ross, 2009) is 98.4305% with execution time 42.77515 second.So the accuracy and speed of iris recognition has improved by the enhanced iris segmentation using optimal minimum cost path dynamic programming (DP) technique.,Master of Information Technology,Certification of Master's / Doctoral Thesis" is not available"
Pages:	133
Call Number:	TA1638.4.N764 2016 3 tesis
Publisher:	UKM, Bangi
Appears in Collections:	Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.