Generative adversarial network for low?light image enhancement

Low-light image enhancement is rapidly gaining research attention due to the increasing demands of extreme visual tasks in various applications. Although numerous methods exist enhance qualities low light, it still undetermined how trade-off between human observation and computer vision processing. In this work, an effective generative adversarial network structure proposed comprising both densely residual block (DRB) enhancing (EB) for low-light enhancement. Specifically, end-to-end method, consisting a generator discriminator, trained using hyper loss function. The DRB adopts dense skip connections connect features extracted from different depths while EB receives unique multi-scale ensure feature diversity. Additionally, sizes allows discriminator further distinguish fake real images patch levels. merits function are also studied recover contextual local details. Extensive experimental results show that our method capable dealing with extremely scenes realistic outperforms several state-of-the-art number qualitative quantitative evaluation tests.