Analysing the Performance of the Algorithms Used to Localize the Iris Region in Eye Images Submitted to Severely Compressed Images

摘要

This work aims to evaluate how does the performance of the iris localization algorithms is influenced by severe compression of eye images used in an automatic iris recognition system. The correct localization of the iris region is essential to guarantee the credibility of the biometric recognition. As it is the first processing stage of the system, it is the first stage that can have its efficiency depredated by the quality loss of the images to be processed. Fractal compression and JPEG2000 compression were applied to a public iris images database. Two traditional methods which localize the circular borders of the iris and a method for detection of eyelids and eyelashes were applied to the compressed images and their efficiencies were evaluated. The results obtained with the simulations, when the images were submitted to several compression levels, lead to the important conclusion that segmentation stage does not represent a barrier for the utilization of compressed images in an iris recognition system.