Scholarlink

<p>Low-light image enhancement (LLIE) is crucial for improving visibility and quality in poorly illuminated conditions, with applications in autonomous driving, surveillance, and medical diagnostics. Traditional methods often introduce artifacts, while deep learning approaches struggle with long-range dependencies. We propose RetinexMemory, a novel framework that integrates a lightweight multi-granularity memory module into RetinexFormer, enhancing both local details and global consistency. Experiments on LOL-v1 and LOL-v2 benchmarks demonstrate competitive performance in PSNR and SSIM while maintaining balanced RMSE. Furthermore, enhanced images also improve downstream tasks such as object detection, confirming the robustness and practical value of RetinexMemory.</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

Enhancing low-light image quality: a multi-granularity memory approach with RetinexMemory

  • 期刊论文

  • 发表日期：

    2026年2月8日

查看全文
  • Xinyang Zhang,
  • Lei Zhang

摘要

Low-light image enhancement (LLIE) is crucial for improving visibility and quality in poorly illuminated conditions, with applications in autonomous driving, surveillance, and medical diagnostics. Traditional methods often introduce artifacts, while deep learning approaches struggle with long-range dependencies. We propose RetinexMemory, a novel framework that integrates a lightweight multi-granularity memory module into RetinexFormer, enhancing both local details and global consistency. Experiments on LOL-v1 and LOL-v2 benchmarks demonstrate competitive performance in PSNR and SSIM while maintaining balanced RMSE. Furthermore, enhanced images also improve downstream tasks such as object detection, confirming the robustness and practical value of RetinexMemory.