Abstract

In recent years, there has been a meteoric rise in the application of deep neural networks for the purpose of iris segmentation. This can be attributed to the extraordinary capacity for learning possessed by the convolution kernels that are utilised by CNNs. Conventional methods have several drawbacks, some of which can be partially compensated for by using CNN-based algorithms, which increase the segmentation precision. On the other hand, the CNN-based iris segmentation approaches that are currently in use typically require a more complex network, which results in an increase in the number of parameters. This is essential to realise a higher degree of precision in the results. CNN-based techniques are effective, they can only be used for a specific application. This makes them inappropriate for general iris segmentation goals, even though they are effective.