Scholarlink

<p>Edge-preserving image smoothing is an essential tool for image processing and computer graphics, as it can be widely applied to various tasks. However, different tasks may favor distinct edge-preserving capabilities. Most existing filters fail to consider the task-dependent smoothing behavior, thus may be subjective to various visually distracting artifacts. In this paper, aiming at improving the versatility, we propose a generalized framework for edge-preserving smoothing based on the isotropic Welsch/Leclerc regularization. We show that the proposed filter is able to provide multi-sparsity gradient regularization, thus enabling varying edge-preserving capabilities. Thanks to the generality, our filter facilitates a variety of tasks. To solve the proposed model, we propose an efficient algorithm based on the additive half quadratic method and Fourier domain optimization. We have conducted both quantitative and qualitative experiments to evaluate the proposed filter. Experimental results demonstrate the superior performance of the proposed filter on several popular low-level vision tasks over the state-of-the-art methods. Furthermore, our filter is highly efficient. It is able to process 720P RGB images in real time on a modern GPU.</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

Multi-sparsity isotropic gradient regularization for edge-preserving image smoothing

  • 期刊论文

  • 发表日期：

    2026年1月8日

查看全文
  • Yang Yang,
  • Xi Zhang,
  • Yiwen Xiong,
  • Xinsheng Wang,
  • Xiangyu Leng

摘要

Edge-preserving image smoothing is an essential tool for image processing and computer graphics, as it can be widely applied to various tasks. However, different tasks may favor distinct edge-preserving capabilities. Most existing filters fail to consider the task-dependent smoothing behavior, thus may be subjective to various visually distracting artifacts. In this paper, aiming at improving the versatility, we propose a generalized framework for edge-preserving smoothing based on the isotropic Welsch/Leclerc regularization. We show that the proposed filter is able to provide multi-sparsity gradient regularization, thus enabling varying edge-preserving capabilities. Thanks to the generality, our filter facilitates a variety of tasks. To solve the proposed model, we propose an efficient algorithm based on the additive half quadratic method and Fourier domain optimization. We have conducted both quantitative and qualitative experiments to evaluate the proposed filter. Experimental results demonstrate the superior performance of the proposed filter on several popular low-level vision tasks over the state-of-the-art methods. Furthermore, our filter is highly efficient. It is able to process 720P RGB images in real time on a modern GPU.