基于Retinex的低光内窥镜图像增强网络  

Low-light Endoscopic Image Enhancement Network Based on Retinex

作　　者：黄东晋[1] 曾子洋 刘金华 马雨奇 

机构地区：[1]上海大学电影学院,上海200072

出　　处：《工业控制计算机》2025年第1期121-122,125,共3页Industrial Control Computer

摘　　要：内窥镜图像往往存在亮度过低、纹理不清晰等问题,因此通过低光图像增强技术来对其进行增强是非常必要的。针对这一任务,提出了一种基于Retinex理论的低光内窥镜图像增强方法,设计了一个新的增强网络及相应的损失函数,并引入CLAHE先验;此外,还创建了一个拥有成对高质/低质图像的内窥镜图像数据集。实验结果表明,该方法能够有效提升内窥镜图像的质量,在三项指标上均优于现有的先进低光图像增强方法。Endoscopic images often have problems such as low brightness and unclear texture,so it is necessary to enhance them by low-light image enhancement techniques.To address this task,this paper proposes a low-light endoscopic image enhancement method based on Retinex theory,designs a new enhancement network and the corresponding loss function,and introduces the CLAHE prior.In addition,an endoscopic image dataset with paired high-quality/low-quality images is created.The experimental results show that the method can effectively improve the quality of endoscopic images,and outperforms the existing state-of-the-art low-light image enhancement methods in all three metrics.

关 键 词：RETINEX理论 内窥镜图像 图像增强 监督学习 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]