In the era of information and machine learning, low-light enhancement technology is a growing research topic.
The development of a low-light enhancement algorithm will improve the overall contrast of an image while restoring color and texture details to obtain more accurate and distinct scene information.
In the study, published in Frontiers of Optoelectronics, researchers from Huazhong University of Science and Technology (HUST) have developed a low-light enhancing algorithm that effectively balances speed with enhancement performance.
The neural network-based algorithm first brightens the low-light images before correcting degradation factors through a two-stage network, a process that was inspired by the processing approach of digital cameras.
During their testing, the team demonstrated that their algorithm could achieve improved performance when compared to modern state-of-the-art methodologies. Additionally, the algorithm maintains a small model size while offering excellent scene adaptability.