Matching thermal spectrum face images against visible spectrum face images has received increased attention in the literature, due to its broad applications in the military, commercial, and law enforcement domains. Thermal emissions from the face images are less sensitive to changes in ambient lighting. Further, thermal face images can be acquired in dark environments characterized by low ambient lighting thereby making them suitable for nighttime face recognition. This project aims to leverage recent advances in generative adversarial networks to synthesize high-fidelity visible face images from the thermal counterparts, as well as explain the process of thermal-to-visible image translation.
The idea candidate should have strong skills in both mathematics and programming. Some knowledge in image/video processing, computer vision and deep learning are preferred.