Deep Lightening Network for Low-Light Image Enhancement

We propose a Deep Lightening Network (DLN) for low-light image enhancement. Inspire by the domain transfer study, we propose a novel cycle learning structure to learn the mapping relationship between low- and normal-light images. Each DLN consists of several Lightening Back-Projection (LBP) blocks that learn the residual between low- and normal-light images. To efficiently estimate the local and global information, we fuse the features from different LBP results. Experimental results on different datasets show that our proposed DLN approach outperforms other approaches in all objective and subjective measures.

I. Introduction

Nowadays, people often take photos to record memorial moments of daily life. However, ample illumination is needed for a photo with good quality. Images captured with insufficient illumination make it difficult for objects/scene recognition. Pixels of a low-light image are always in a low-dynamic range, which makes the image be too dim. There are some possible solutions to solve this problem: 1) To take photos with a higher ISO sensitivity: It can improve the visibility of the dark area. Yet, it usually causes more noise and overexposes to the normal-light areas. 2) To use flash: This is an effective solution. But it may bother others and is not allowed in some places (e.g. museum, art gallery, etc.). 3) To use longer exposure time: It may suffer from blur problems and is not suitable for shooting videos (the interval of video frames may be too short).