Facial Images Quality Assessment based on ISO/ICAO Standard Compliance Estimation by HMAX Model

Facial images are most popular biometrics in automated systems. Different methods have been introduced to evaluate the quality of these images. FICV is a common benchmark to evaluate ISO / ICAO compliancy Assessment algorithms. In this work a model of brain functionality for Facial Image Quality Assessment, bases on FICV benchmark has been introduced. It is tried to use HMAX model for brain functionality simulation and evaluate its operation. Based on the accuracy of compliancy verification, Equal Error Rate of ICAO requirements, has been classified and from those with higher error rate in the past researches, nine ICAO requirements have been used to assess the compliancy of the face images quality to the standard, in this work. To evaluate the quality of facial images, first, image patches were generated for key and non-key face components by using Viola-Jones algorithm. For brain function simulating, HMAX method has been applied to these patches. In the HMAX model, a multi-resolution spatial pooling has been used which encodes local and public spatial information for generating image discriminative signatures. In the proposed model, the way of storing and fetching information is similar to the function of the brain. For training and testing the model, AR and PUT databases were used. The results has been evaluated by FICV assessment factors and show lower Equal Error Rate and rejection rate, compared to existing methods, in the evaluation of some requirements, especially in the cases of Frame Too Heavy (ICAO19) and Frame Across Eyes (ICAO20).