Scholarlink

<p>Shadows in images, caused by objects blocking light, degrade image quality and complicate various computer vision tasks. This paper presents a novel two-stage shadow removal method that addresses this challenge through separate shadow detection and recovery stages. In the first stage, the GLFFNet model leverages global and local channel fusion modules to precisely detect shadow regions. In the second stage, the USFNet model employs shadow localization, boundary enhancement, and frequency band extraction modules to recover shadow areas while minimizing impact on non-shadowed regions. Extensive experiments on multiple datasets demonstrate the superior performance of our method, achieving state-of-the-art results in shadow detection and recovery. This work contributes to advancing shadow removal techniques for practical computer vision applications. The code and models used in this paper are available at: <a href="https://github.com/Valsemia/GLFFNet-and-USFNet.git">https://github.com/Valsemia/GLFFNet-and-USFNet.git</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

A two-stage shadow removal method with multi-channel multi-scale feature fusion and stripe pooling attention

  • 期刊论文

  • 发表日期：

    2025年9月30日

查看全文
  • Zhidan Li,
  • Zhou Chen,
  • Jiayue Xu,
  • Jixiang Cheng

摘要

Shadows in images, caused by objects blocking light, degrade image quality and complicate various computer vision tasks. This paper presents a novel two-stage shadow removal method that addresses this challenge through separate shadow detection and recovery stages. In the first stage, the GLFFNet model leverages global and local channel fusion modules to precisely detect shadow regions. In the second stage, the USFNet model employs shadow localization, boundary enhancement, and frequency band extraction modules to recover shadow areas while minimizing impact on non-shadowed regions. Extensive experiments on multiple datasets demonstrate the superior performance of our method, achieving state-of-the-art results in shadow detection and recovery. This work contributes to advancing shadow removal techniques for practical computer vision applications. The code and models used in this paper are available at: https://github.com/Valsemia/GLFFNet-and-USFNet.git.