对比度(视觉)
计算机科学
人工智能
计算机视觉
亮度
照度
伽马校正
规范化(社会学)
联营
对比度增强
图像增强
图像(数学)
光学
物理
社会学
磁共振成像
放射科
医学
人类学
作者
Zhaorun Zhou,Zhongchao Shi,Wenqi Ren
出处
期刊:IEEE Transactions on Instrumentation and Measurement
[Institute of Electrical and Electronics Engineers]
日期:2023-01-01
卷期号:72: 1-16
被引量:2
标识
DOI:10.1109/tim.2022.3232641
摘要
Images captured under low-illumination conditions usually suffer from severe degradations, such as fading and low contrast, drastically affecting the performance of systems relying on images under low-illumination conditions. To address such problems, this study proposes a linear contrast enhancement network (LCENet) for low-illumination image enhancement. It consists of three subnets: two encoder–decoder-based subnets for gradient map restoration and brightness enhancement, respectively, and a backbone network for adaptive brightness and contrast adjustment. In addition, a linear contrast enhancement adaptive instance normalization (LCEAIN) module with linear contrast enhancement ability is proposed in the backbone network, which can avoid the problem of ignoring contrast enhancement when enhancing image brightness. Considerable evaluations on both synthetic and real low-illumination images show that the proposed method performs favorably against other existing similar methods. Moreover, our method can handle complex low-illuminance conditions and has good generalization for low-illuminance scenes with backlighting, night scenes with light sources, as well as underwater scenes with low illuminance. Code: https://github.com/zhouzhaorun/LCENet .
科研通智能强力驱动
Strongly Powered by AbleSci AI