LSD$_2$ - Joint Denoising and Deblurring of Short and Long Exposure Images with Convolutional Neural Networks

This paper addresses the challenging problem of acquiring high-quality photographs with handheld smartphone cameras in low-light imaging conditions. We propose an approach based on capturing pairs of short and long exposure images in rapid succession and fusing them into a single high-quality photograph using a convolutional neural network. The network input consists of a pair of images, where the short exposure image is typically noisy and has poor colors due to low lighting and the long exposure image is susceptible to motion blur when the camera or scene objects are moving. The network is trained using a combination of real and simulated data and we propose a novel approach for generating realistic synthetic short-long exposure image pairs. Our approach is the first one to address the joint denoising and deblurring problem using deep networks. It outperforms the existing denoising and deblurring methods in this task and allows to produce good images in extremely challenging conditions. Our source code, pretrained models and data will be made publicly available to facilitate future research.