Real-time iris image quality evaluation implemented in Ultrascale MPSoC

Iris recognition consists of measuring the unique patterns of the coloured and textured part of the eye (the iris), and comparing the obtained "signature" with those of a dataset to verify and authenticate identity. The uniqueness of the iris pattern makes the technique more robust and accurate for verification compared to other biometric systems. One of the main challenges of iris recognition is the ability to work with people on the move and at an increasing distance (over 1 metre). While the first implementations of iris recognition require the eyes to be placed at a close distance to the sensor and to maintain a static position during eye detection and capture, nowadays the iris recognition technique is progressing to be less intrusive. In this scenario, the problem of detecting the iris with an adequate quality (contrast, resolution) to be useful to identify the person, can force the framework to be able to process a large amount of frames per second (fps). This article describes the implementation in an MPSoC (Multiprocessor System-on-Chip) of a core that allows an eye detection framework to quickly evaluate the quality of the captured iris images. Thus, it is possible to remove the poor quality images from the subsequent processing steps. This implementation is intended to discard invalid images for person recognition therefore reducing the load on other parts of the system. It has been successfully integrated into an eye detection framework capable of processing over 50 fps working with a 16 Mpixel sensor.