Scholarlink

<p>Edge-preserving image smoothing is fundamental in the fields of computer vision and image processing. The primary challenge is to smooth low-amplitude details while retaining critical structural information. Existing global filtering methods typically incorporate a data fidelity term and a gradient smoothness term. However, preserving the full semantic information of an image solely through data fidelity remains difficult. To address this, we propose a novel generalized smoothing model that integrates a data fidelity term, a structural fidelity term, and a sparse smoothing term to enhance edge-preserving smoothing performance. The structural fidelity term is designed to ensure that the gradient of the output image closely matches the preprocessed gradient of the input image, thereby achieving structural fidelity. Simultaneously, the smoothing term with sparse regularity constraints is employed to smooth detailed information while preserving significant structural elements. Extensive experimental validation demonstrates that our proposed method outperforms existing techniques and is applicable across various fields, including image smoothing, detail enhancement, edge extraction, HDR tone mapping, clip-art compression artifact removal, and image abstraction. The source code is available at: <a href="https://github.com/kxZhang1016/EPSGEF">https://github.com/kxZhang1016/EPSGEF</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

Edge-preserving image smoothing via sparse gradient enhancement

  • 期刊论文

  • 发表日期：

    2025年8月29日

查看全文
  • Jianwu Long,
  • Kaixin Zhang,
  • Yuanqin Liu,
  • Shuang Chen,
  • Qi Luo

摘要

Edge-preserving image smoothing is fundamental in the fields of computer vision and image processing. The primary challenge is to smooth low-amplitude details while retaining critical structural information. Existing global filtering methods typically incorporate a data fidelity term and a gradient smoothness term. However, preserving the full semantic information of an image solely through data fidelity remains difficult. To address this, we propose a novel generalized smoothing model that integrates a data fidelity term, a structural fidelity term, and a sparse smoothing term to enhance edge-preserving smoothing performance. The structural fidelity term is designed to ensure that the gradient of the output image closely matches the preprocessed gradient of the input image, thereby achieving structural fidelity. Simultaneously, the smoothing term with sparse regularity constraints is employed to smooth detailed information while preserving significant structural elements. Extensive experimental validation demonstrates that our proposed method outperforms existing techniques and is applicable across various fields, including image smoothing, detail enhancement, edge extraction, HDR tone mapping, clip-art compression artifact removal, and image abstraction. The source code is available at: https://github.com/kxZhang1016/EPSGEF.