الفهرس 
Iris RecognitionOnly 14 pages are availabe for public view
 |  | from | 102 |  |  |
| | | from | 102 | | |
Abstract
Iris localization is an important step in the iris recognition systems. All subsequent steps including iris normalization, feature extraction and matching depend on iris localization accuracy. Traditional iris localization methods often involve an exhaustive search of a threedimensional parameter space; iris center coordinates (x0,y0) and iris radius (r) , which is a time consuming process. This thesis presents a comparative study between the most three common iris localization algorithms: Integro-differential operator, Masek algorithm and Distance Regularized Level Set Evolution (DRLSE). This comparative study is performed in three cases: normal images, noisy images, and blurred images to know which algorithm resists the different degradation effects. On the other hand, we present an algorithm for iris recognition based on deep learning. This algorithm depends on Convolutional Neural Networks (CNNs), and it manages to get an accuracy of recognition up to 100% on Twins database.