Cross-spectral face recognition (CFR) is aimed at recognizing individuals, where compared face images stem from different sensing modalities, for example infrared vs. visible. While CFR is inherently more challenging than classical face recognition due to significant variation in facial appearance associated to a modality gap, it is superior in scenarios with limited or challenging illumination, as well as in the presence of presentation attacks. Recent advances in artificial intelligence related to convolutional neural networks (CNNs) have brought to the fore a significant performance improvement in CFR. Motivated by this, the contributions of this survey are three-fold. We provide an overview of CFR, targeted to compare face images captured in different spectra, by firstly formalizing CFR and then presenting concrete related applications. Secondly, we explore suitable spectral bands for recognition and discuss recent CFR-methods, placing emphasis on deep neural networks. In particular we revisit techniques that have been proposed to extract and compare heterogeneous features, as well as datasets. We enumerate strengths and limitations of different spectra and associated algorithms. Finally, we discuss research challenges and future lines of research.