Comprehensive Study of Low Light Facial Recognition Based on Conventional and Deep Learning Models

Facial recognition is a widely used biometric system for detecting or verifying individuals based on their facial features. This technique identifies a person's emotions by analysing their facial images using Artificial Intelligence (AI) for broad applications. This application is employed in security systems, surveillance, access control, and personal devices. The technique operates steadily in ideal lighting conditions but the performance decreases under low-light facial environments due to the presence of noise and poor visibility in input images. Low-light facial recognition plays a vital role in research to ensure reliable identification in real-world conditions for effectively recognizing facial expressions, even with low resolution images. Various conventional models and Deep Learning (DL) models have been introduced for this purpose. Conventional and DL models showcase promising results in enhancing image quality and the model's recognition accuracy. This survey presents an overview of recent conventional and DL approaches for low-light facial recognition, comparing the existing models by discussing their effectiveness, highlighting their limitations and directing future research in this field.