基于深度学习的低光照图像增强研究综述OA北大核心

The aim of low-light image enhancement is to optimize images captured in low-light environments by improving their brightness and contrast.Currently,deep learning has become the main method in the field of low-light image enhancement,necessitating a review of deep learning-based methods.First,this paper classified traditional methods of low-light image en-hancement and analyzed and summarized their advantages and disadvantages.Then,this paper focused on deep learning-based methods,classified them into supervised and unsupervised categories,and summarized their respective advantages and disad-vantages.This paper also summarized the loss functions applied in deep learning approaches.Next,this paper briefly summa-rized the commonly used datasets and evaluation metrics,using information entropy to quantitatively compare traditional me-thods,and employing peak signal-to-noise ratio and structural similarity to objectively evaluate deep learning-based methods.Finally,this paper summarized the shortcomings of current methods and prospect future research directions.