Scholarlink

The integration of CIDNet with the Mamba algorithm and ResBlock presents an effective low-light image enhancement (LLIE) method. CIDNet’s dual-branch structure, comprising HV and I branches in the HVI color space, processes color and illumination components separately while enabling information exchange via the LCA module. This design preserves color naturalness under varying lighting. The Mamba algorithm adapts to illumination changes, while ResBlock enhances feature extraction and integration. Their fusion within CIDNet ensures coherent image enhancement. Experiments on LOL datasets highlight superior performance, with PSNR improvements of 4.22% over Rentinexmamba and 3.72% over MIRNet on LOLv1. The method balances quality and naturalness, demonstrating robustness against low-light challenges.

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 Enhancement with Mamba-Integrated Dual-Branch Neural Networks

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Enze Pan

摘要

The integration of CIDNet with the Mamba algorithm and ResBlock presents an effective low-light image enhancement (LLIE) method. CIDNet’s dual-branch structure, comprising HV and I branches in the HVI color space, processes color and illumination components separately while enabling information exchange via the LCA module. This design preserves color naturalness under varying lighting. The Mamba algorithm adapts to illumination changes, while ResBlock enhances feature extraction and integration. Their fusion within CIDNet ensures coherent image enhancement. Experiments on LOL datasets highlight superior performance, with PSNR improvements of 4.22% over Rentinexmamba and 3.72% over MIRNet on LOLv1. The method balances quality and naturalness, demonstrating robustness against low-light challenges.