Scholarlink

<p>Underwater images often suffer from severe color distortion and low visibility caused by wavelength-dependent light absorption and scattering. This work presents a hybrid enhancement framework that integrates a Contrastive Cascaded Learning Network (CCL-Net) with an adaptive Guided Filter (GF) to recover natural brightness and edge definition. The CLAHE–UCM fusion method is used to generate pseudo-reference images for supervised training, ensuring color fidelity and illumination consistency. The proposed model achieves 35.04&#xa0;dB PSNR, 0.923 SSIM, and higher UIQM/UCIQE scores than existing approaches, confirming perceptually balanced restoration. Enhanced outputs substantially improve YOLOv11-based debris detection, increasing mean Average Precision by 6.2% over raw images. The results demonstrate that the joint CCL-Net + GF pipeline provides an efficient solution for real-time underwater vision tasks.</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

CCL-guided EdgeNet: an edge-preserving deep learning framework for low-light underwater image enhancement

  • 期刊论文

  • 发表日期：

    2026年1月21日

查看全文
  • S. Mourish Kanna,
  • S. K. Shiva Ram Ganesh,
  • V. Hariharan,
  • H. Kavietha,
  • L. K. Pavithra

摘要

Underwater images often suffer from severe color distortion and low visibility caused by wavelength-dependent light absorption and scattering. This work presents a hybrid enhancement framework that integrates a Contrastive Cascaded Learning Network (CCL-Net) with an adaptive Guided Filter (GF) to recover natural brightness and edge definition. The CLAHE–UCM fusion method is used to generate pseudo-reference images for supervised training, ensuring color fidelity and illumination consistency. The proposed model achieves 35.04 dB PSNR, 0.923 SSIM, and higher UIQM/UCIQE scores than existing approaches, confirming perceptually balanced restoration. Enhanced outputs substantially improve YOLOv11-based debris detection, increasing mean Average Precision by 6.2% over raw images. The results demonstrate that the joint CCL-Net + GF pipeline provides an efficient solution for real-time underwater vision tasks.