Scholarlink

<p>Image inpainting, a pivotal technique in computer vision, aims to fill missing regions with semantically and visually coherent pixels. This paper introduces Enhanced LW-painter, a lightweight image inpainting network designed to balance performance and model efficiency. Enhanced LW-painter employs an enhanced large kernel module to expand the receptive field and a feature fusion module to enhance feature expressiveness. Experiments on the FFHQ, LFW, Dunhuang Mogao Grottoes Mural, and Paris Streetview datasets demonstrate that Enhanced LW-painter effectively inpaints randomly occluded images, generating detailed and semantically accurate results with limited computational resources.</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

Enhanced LW-painter: Lightweight Image Inpainting via Large Receptive Field and Feature Fusion Optimization

  • 期刊论文

  • 发表日期：

    2025年9月9日

查看全文
  • Sihong Yang,
  • Qian Zhang,
  • Yucheng Yang,
  • Jiliang Shi,
  • Wuer Bai,
  • Shuang Liu

摘要

Image inpainting, a pivotal technique in computer vision, aims to fill missing regions with semantically and visually coherent pixels. This paper introduces Enhanced LW-painter, a lightweight image inpainting network designed to balance performance and model efficiency. Enhanced LW-painter employs an enhanced large kernel module to expand the receptive field and a feature fusion module to enhance feature expressiveness. Experiments on the FFHQ, LFW, Dunhuang Mogao Grottoes Mural, and Paris Streetview datasets demonstrate that Enhanced LW-painter effectively inpaints randomly occluded images, generating detailed and semantically accurate results with limited computational resources.