Scholarlink

<p>With the development of modern urban systems, the use of cameras for video surveillance of urban environments has become an essential requirement. However, in nighttime scenes, especially under foggy weather conditions, the visual quality can significantly degrade due to uneven atmospheric illumination, leading to color distortion and errors in transmission rate estimation. In this study, we propose a novel nighttime dehazing algorithm that integrates Retinex theory with the light–dark channel prior. Specifically, (1) we introduce a Retinex-based variational model to estimate global atmospheric light under low illumination conditions, effectively correcting color biases in the dehazed images; and (2) we combine the light and dark channel priors to refine the transmission rate estimation, resulting in an optimized dehazing framework. Extensive experiments on a real-world nighttime dataset demonstrate the method's applicability to varying fog densities and complex light source distributions. Experimental results show significant improvements in both color fidelity and detail preservation, enhancing the reliability of urban infrastructure video surveillance systems in challenging nighttime foggy environments.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Nighttime image dehazing via Retinex theory and bright-dark channel priors

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月6日

查看全文
  • Xianglei Liu,
  • Yahao Wu,
  • Runjie Wang,
  • Yuhang Liu

摘要

With the development of modern urban systems, the use of cameras for video surveillance of urban environments has become an essential requirement. However, in nighttime scenes, especially under foggy weather conditions, the visual quality can significantly degrade due to uneven atmospheric illumination, leading to color distortion and errors in transmission rate estimation. In this study, we propose a novel nighttime dehazing algorithm that integrates Retinex theory with the light–dark channel prior. Specifically, (1) we introduce a Retinex-based variational model to estimate global atmospheric light under low illumination conditions, effectively correcting color biases in the dehazed images; and (2) we combine the light and dark channel priors to refine the transmission rate estimation, resulting in an optimized dehazing framework. Extensive experiments on a real-world nighttime dataset demonstrate the method's applicability to varying fog densities and complex light source distributions. Experimental results show significant improvements in both color fidelity and detail preservation, enhancing the reliability of urban infrastructure video surveillance systems in challenging nighttime foggy environments.