Software implementation of the OSIRIS iris recognition algorithm in FPGA

Abstract

International audienceIn this paper, we present a software implementation of an iris recognition algorithm on a FPGA (Field Programmable Gate Array). We use the OSIRIS reference systems developed in the framework of the BioSecure project which is an implementation of the well-knoun Daugman approach [10]. This system is composed of two modules. The first one uses the Hough transform to segment and normalize the iris. The second module applies a Gabor filter on the normalized image and calculates the similarity score between two images using the Hamming distance. OSIRIS algorithm has performance (EER=7.04%) when it is running on the PC in ICE 2005 Database. This system is coded in C++ using the library OpenCV (Open Source Computer Vision). This library is difficult to implement it in FPGA. that's why, we first remove this library in our code than improve this algorithm. Finally, we implement it on a CYCLONE II DE2 Board FPGA using the NIOS II processor which is characterized by its flexibility and programmability

Iris Matching Step Implementation in FPGA

“…In 2011, [8] authors implemented iris recognition system in software using C++ language, then they improved the system and programmed it in VHDL language and synthesized using CycloneII EP2C35FPGA. The improved system consists of two modules, first module implemented using Hough Transform for segmentation and normalization steps, while the second used Gabor filter on the normalized iris and Hamming distance for calculating the similarity between images.…”