Deep Learning Techniques for Face Image Quality Estimation

Abstract

Biometric authentication using fingerprints or face recognition is becoming mainstream, and there is a need to make these methods as secure and reliable as possible. One way to achieve better performance with a biometric authentication method, is to introduce a quality estimation step early in the pipeline, so that low-quality images can be discarded, and a new image acquired.

This thesis investigates the possibility of using deep learning techniques for doing face image quality estimation on a smartphone. We make use of different types of LSTM (Long Short-Term Memory) neural networks and find that they perform well compared to an existing commercial solution, and also compared to other types of neural network.

The result of our work is an iOS framework that can be used perform image quality analysis. This framework can be easily integrated into any iPhone app in order to strengthen its face recognition capabilities.