Scholarlink

<p>To address the issues of overall darkness and low contrast in surveillance video images caused by complex underground coal mine spaces and uneven illumination, this study proposes a multi-scale enhancement algorithm named GCA for low-illumination coal mine images. The algorithm is based on Contrast Limited Adaptive Histogram Equalization (CLAHE) and Automatic Color Equalization (ACE). First, a CLAHE algorithm integrated with dynamic Gamma correction (GC-CLAHE) is introduced. Traditional contrast-limited histogram equalization is enhanced by adopting a grid search-based weighted averaging algorithm to adaptively adjust Gamma parameters, thereby alleviating contrast imbalance caused by local underexposure or overexposure in underground images. Second, the local gradient-guided enhancement mechanism of the ACE algorithm is incorporated to strengthen textural details while suppressing noise, achieving collaborative optimization of luminance balance and edge features. Finally, the processed images are weighted and fused to generate the final enhanced result. Comparative and ablation experiments are conducted using the CUMT-CMUID dataset, with Single-Scale Retinex (SSR), Multi-Scale Retinex (MSR), Dark Channel Prior (DCP), Guided Filtering, Homomorphic Filtering and adaptive sigmoid transfer function (ASTF) as the comparison enhancement methods. Both subjective and objective evaluations are performed on the enhanced images, using information entropy, mean square error, standard deviation, and average gradient as objective metrics. Experimental results show that the proposed GCA algorithm achieves Information Entropy, Standard Deviation, and Mean Gradient values of 7.758, 55.430, and 80.592, respectively. Compared with SSR, these values represent improvements of 24.64%, 13.09%, and 121.97%; compared with MSR, the improvements are 20.48%, 29.95%, and 202.45%. The algorithm outperforms the comparison methods significantly in multiple evaluation metrics. It effectively enhances image brightness while preserving the natural appearance of images, improves detail information and contrast, and thus provides a high-reliability preprocessing solution for underground intelligent surveillance and target recognition.</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

A new multi-scale adaptive enhancement algorithm for low-illumination coal mine images based on improved CLAHE and ACE

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月29日

查看全文
  • Dengcong Mu,
  • Ziteng Wang,
  • Zheng Li,
  • Fei Dong

摘要

To address the issues of overall darkness and low contrast in surveillance video images caused by complex underground coal mine spaces and uneven illumination, this study proposes a multi-scale enhancement algorithm named GCA for low-illumination coal mine images. The algorithm is based on Contrast Limited Adaptive Histogram Equalization (CLAHE) and Automatic Color Equalization (ACE). First, a CLAHE algorithm integrated with dynamic Gamma correction (GC-CLAHE) is introduced. Traditional contrast-limited histogram equalization is enhanced by adopting a grid search-based weighted averaging algorithm to adaptively adjust Gamma parameters, thereby alleviating contrast imbalance caused by local underexposure or overexposure in underground images. Second, the local gradient-guided enhancement mechanism of the ACE algorithm is incorporated to strengthen textural details while suppressing noise, achieving collaborative optimization of luminance balance and edge features. Finally, the processed images are weighted and fused to generate the final enhanced result. Comparative and ablation experiments are conducted using the CUMT-CMUID dataset, with Single-Scale Retinex (SSR), Multi-Scale Retinex (MSR), Dark Channel Prior (DCP), Guided Filtering, Homomorphic Filtering and adaptive sigmoid transfer function (ASTF) as the comparison enhancement methods. Both subjective and objective evaluations are performed on the enhanced images, using information entropy, mean square error, standard deviation, and average gradient as objective metrics. Experimental results show that the proposed GCA algorithm achieves Information Entropy, Standard Deviation, and Mean Gradient values of 7.758, 55.430, and 80.592, respectively. Compared with SSR, these values represent improvements of 24.64%, 13.09%, and 121.97%; compared with MSR, the improvements are 20.48%, 29.95%, and 202.45%. The algorithm outperforms the comparison methods significantly in multiple evaluation metrics. It effectively enhances image brightness while preserving the natural appearance of images, improves detail information and contrast, and thus provides a high-reliability preprocessing solution for underground intelligent surveillance and target recognition.