Super-Resolution Method Based on Multiple Multi-Layer Perceptrons for Iris Recognition

Iris recognition is a form of biometric techniques that identifies user with the unique iris patterns between the pupil and the sclera. There is an advantage that its recognition accuracy is very high due to high DOF (degree of freedom) of the iris patterns. However, it requires an iris image with high quality in order to preserve the high recognition accuracy. So, the conventional iris recognition systems have used expensive cameras with zoom lens because an iris image which is obtained from the camera without zoom lens has low resolution at a depth of focus region. To overcome this problem, we propose a new super-resolution method based on multiple MLPs (multilayer perceptrons) for the iris recognition. The proposed method uses three MLPs considering the edge directions of the iris patterns. We produce a high resolution iris image from a single low resolution iris image by using bilinear interpolation based on the output pixel values of the three MLPs. Experimental results showed that the improvement of the recognition accuracy was obtained by using the proposed method.