CEL-Net: Continuous Exposure for Extreme Low-Light Imaging

Michael Klyuchka
Evgeny Hershkovitch Neiterman
Gil Ben-Artzi

Computer Science Department, Ariel University, Israel

arXiv


Code+Dataset
[will be uploaded after publication]





Tune optimal input and output exposures


Output 1


Output 2


Output 3



Highlights


Abstract

Deep learning methods for enhancing dark images learn a mapping from input images to output images with pre-determined discrete exposure levels. At inference time the input and optimal output exposure levels of the given image are often different from the seen ones during training and as a result the enhanced image might suffer from visual distortions, such as low contrast or dark areas. We address this issue by introducing a deep learning model that can continuously generalize at inference time to unseen exposure levels without the need to retrain the model. To this end, we introduce a dataset of 1500 raw images captured in both outdoor and indoor scenes, with five different exposure levels and various camera parameters. Using the dataset, we develop a model for extreme low-light imaging that can continuously tune the input or output exposure level of the image to an unseen one. We investigate the properties of our model and validate its performance, showing promising results.

Network Architecture - Brightness and ISP Tuning Parameters


                     


One Trained Model - Many Enhancement Options at Inference Time



The Dataset

The top two rows are images from outdoor scenes and the bottom two rows are images from indoor scenes.
Left to right are exposure times of 0.1s,0.5s,1s,5s,10s.



                     


Paper


CEL-Net: Continuous Exposure for Extreme Low-Light Imaging,
M. Klyuchka, E.H Neiterman, G. Ben-Artzi
arXiv