Investigating the effects of degradation and restoration on face recognition in under-display cameras

Under-display camera (UDC) technology integrates the camera beneath the screen, enabling a full-screen display without a notch or punch-hole design. However, inevitable degradation occurs due to diffraction caused by display pixels, leading to reduced transmittance, noise, blur, and flare. These optical distortions degrade facial recognition accuracy, limiting the practicality of UDC systems. Despite ongoing research on UDC image restoration, studies on facial recognition performance with UDC remain limited. While a few studies have attempted to generate UDC-distorted images containing human faces, they overlook actual UDC degradations and the pair-matching task, a key smartphone application. This paper presents three UDC facial recognition datasets generated by a generative adversarial network (GAN) trained on synthetic and real-world datasets with flares. We aim to analyze the impact of UDC restoration on face recognition accuracy, particularly in pair matching, contributing to the expansion of UDC face recognition research.