Low-Light Image Enhancement: A Comparative Review and Prospects

Low-light image enhancement is a key prerequisite for diverse applications in the field of image processing and computer vision. Various approaches for this task have been introduced over last few decades, and the current state of the art methods have shown remarkable advances based on deep neural networks. However, there are still technical issues to be resolved, e.g., dependency on subjective re-touching results and inconsistency with subjective evaluations. The goal of this work is to provide a comprehensive overview and a practical guide for experts as well as beginners. This paper covers a systematic taxonomy of existing algorithms, representative methodologies, and the performance analysis on benchmark datasets. To pave the way of the development direction for low-light image enhancement, constructive discussions and prospects are also provided.