Real-time Enhancement Algorithm of Low-light Image Based on Zero-DCE

At present, all walks of life, such as industry, transportation, medical care, etc., can use computer vision technology to promote the development of the industry and solve the problems of the industry. However, some industries have complex operation scenarios and are easily affected by weather and other factors, resulting in blurred and dimmed images. In response to such problems, many image enhancement algorithms have emerged and achieved good results. However, most of these methods perform image enhancement on all video images, which leads to deliberate enhancement of some normal images, resulting in overexposure; at the same time, it also causes waste of computing resources, and it is difficult to guarantee the real-time performance of the algorithm. In view of the above content, this paper first proposes a low-light image judgment method to judge whether the current video frame needs to be preprocessed. In addition, through experiments and comparison of the current mainstream low-light image preprocessing algorithms, the Zero-DeE algorithm that is more suitable for field use is selected, and the self-made drilling multi-exposure sequence data set is combined to train the optimization model, which has both subjective processing effect and objective index evaluation raised.