摘要
针对微光图像存在的低亮度、低对比度和噪声等退化因素对高级视觉任务的影响,本文提出一种基于深度多层次特征提取的微光图像增强方法。通过深度多层次地提取微光图像特征,并使用通道自注意力机制计算通道权重来提升网络颜色恢复的能力。将非线性激活的特征图送入不共享参数的特征增强模块,并对不同层次的特征分别增强来提升网络处理不同退化因素的能力。实验结果表明,本文方法能同时处理微光图像的各种退化因素,在主观评价和多种客观评价指标上,均优于其他对比方法。
This paper proposes a low light image enhancement method based on deep multi-level feature extraction to address the impact of degradation factors such as low brightness,low contrast,and noise on advanced visual tasks in low light images.By extracting low light image features at multiple levels of depth and using channel self attention mechanism to calculate channel weights,the ability of network color restoration is improved.Send the non-linear activated feature map into a feature enhancement module that does not share parameters,and enhance different levels of features separately to improve the network's ability to handle different degradation factors.The experimental results show that the proposed method can simultaneously handle various degradation factors of low light images,and is superior to other comparative methods in subjective evaluation and multiple objective evaluation indicators.
作者
吴振才
崔世杰
WU Zhencai;CUI Shijie(Department of Information and Electronic Engineering,Shangqiu Institute of Technology,Shangqiu,China,476000)
出处
《福建电脑》
2025年第5期13-17,共5页
Journal of Fujian Computer
关键词
图像增强
多层次深度特征
微光图像
Image Enhancement
Multi Level Deep Features
Low Light Image
