Feature Fusion Network for Low-Light Image Enhancement

Abstract

With the popularity of electronic cameras, we can get pictures under various lighting conditions in our daily life. The images which obtained in low illumination environment will lose contrast, colour, and content details which will bring obstacles for understanding the image. To tackle this problem, a method is proposed in this paper to tackle low-light image enhancement. This method fuses the high level and the low level feature, uses an attention module to model the context information and selects the critical feature to supplemented the high level feature to reconstruct the enhanced image. A number of experiments demonstrate that our method can get images with improved and reduced noise.