Abstract
Photography during night or in dark conditions typically suffers from noise,low light and blurring issues due to the dim environment and the common use oflong exposure. Although Deblurring and Low-light Image Enhancement (LLIE) arerelated under these conditions, most approaches in image restoration solvethese tasks separately. In this paper, we present an efficient and robustneural network for multi-task low-light image restoration. Instead of followingthe current tendency of Transformer-based models, we propose new attentionmechanisms to enhance the receptive field of efficient CNNs. Our method reducesthe computational costs in terms of parameters and MAC operations compared toprevious methods. Our model, DarkIR, achieves new state-of-the-art results onthe popular LOLBlur, LOLv2 and Real-LOLBlur datasets, being able to generalizeon real-world night and dark images. Code and models athttps://github.com/cidautai/DarkIR