Low-light Image Enhancement by Edge Attention Guidance and Multi-scale Feature Fusion

Images captured at low-light environment often suffer from low brightness, insufficient contrast and detail loss, which hinders further processing and application of the images. Thus, a low-light image enhancement algorithm based on the edge attention guidance and multi-scale feature fusion is proposed. First, a dense residual network is constructed as a feature extractor, and the extracted feature maps at different scales are fused by using a modified RefineNet, which makes full use of the feature information in the image. Second, an edge attention mechanism is designed to generate an attention map based on the edge detection results. Instead of attention networks, the attention map is combined with the loss function to guide the network training. This way, the edge detail information hidden in the dark is enhanced without increasing the network inference burden as well as the network complexity. Experiments performed on synthetic and low-illumination dataset images show that the proposed method can effectively improve the image brightness and contrast, restore image edge details.