Scholarlink

<p>Coal rock image acquisition often suffers from poor quality due to challenges such as inadequate lighting, noise, and sensor limitations. To address these issues, we introduce ISTFormer, a lightweight transformer-based approach for coal rock image super-resolution. ISTFormer integrates a global permuted self-attention block with a local convolutional block, leveraging iterative up-and-down sampling to extract comprehensive features from low-resolution images. Experiments on benchmark datasets and a custom coal rock dataset, CD-188, demonstrate ISTFormer’s superiority over existing lightweight super-resolution methods, achieving state-of-the-art results with improved PSNR and SSIM scores. By open-sourcing our algorithm’s code and datasets, we aim to foster further developments in this field. This work presents a significant step toward enhancing coal rock image quality for improved identification and classification accuracy. Our source code is available at: <a href="https://github.com/haoliuliuhao/ISTFormer">https://github.com/haoliuliuhao/ISTFormer</a>.</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

ISTFormer: lightweight transformer for enhanced super-resolution of coal rock images via iterative feature extraction

  • 期刊论文

  • 发表日期：

    2025年2月3日

查看全文
  • Hao Liu,
  • Ye Liu,
  • Shuanglong Yao,
  • Tongshuai Yu,
  • Ke Gao,
  • Pengcheng Hao,
  • Shuqing He,
  • Ji Chen,
  • Xing Wang

摘要

Coal rock image acquisition often suffers from poor quality due to challenges such as inadequate lighting, noise, and sensor limitations. To address these issues, we introduce ISTFormer, a lightweight transformer-based approach for coal rock image super-resolution. ISTFormer integrates a global permuted self-attention block with a local convolutional block, leveraging iterative up-and-down sampling to extract comprehensive features from low-resolution images. Experiments on benchmark datasets and a custom coal rock dataset, CD-188, demonstrate ISTFormer’s superiority over existing lightweight super-resolution methods, achieving state-of-the-art results with improved PSNR and SSIM scores. By open-sourcing our algorithm’s code and datasets, we aim to foster further developments in this field. This work presents a significant step toward enhancing coal rock image quality for improved identification and classification accuracy. Our source code is available at: https://github.com/haoliuliuhao/ISTFormer.