Representation a new filter to improve iris segmentation in identity recognition systems

Iris is a suitable biometric to replace present security systems. In this paper, Hough transform is used to iris segmentation. But to reduce noise and increase the quality, the MLP neural network is used instead of the old algorithm, in order to adjust the filter mask, and two-dimensional convolution (2D convolution) is used instead of large amount of calculations for edge detection. For this purpose, Canny filter is selected as the target (objective) function of neural network. In addition to having a high ability to design matrix filters, this structure can also greatly reduce the existing noise. To test this algorithm, the CASIA database is used. The result shows an improvement in the quality of segmented image and an increase in accuracy of system.