Low-Light Image Enhancement Algorithm Based on Multi-scale Concat Convolutional Neural Network

We present a deep learning-based method for lowlight image enhancement. Recent approaches, based on deep neural networks, produce impressive results but are either too slow to run at practical resolutions, or still contain the problems of excessive enhancement, unclear details and performance degradation. To address these tasks, we propose the Multi-scale Concat Image Enhancement Network (MCIEN). The core of our approach is a feed-forward neural network that learns affine transforms of local and global features. By these means, our network is able to recover clear details, distinct contrast, and natural color in the enhancement results. We perform extensive experiments on the benchmark MIT-Adobe FiveK dataset, and show that our network is superior to other contrast algorithms in visual effects and quantitative evaluations.