Scholarlink

<p>Current research on super-resolution (SR) still suffer from undesirable structural distortions, limiting their effectiveness in preserving critical edge and texture details. To address this issue, we propose incorporating gradients as high-frequency information to enhance structural preservation. Building on this idea, we introduce the Swin Transformer-based Gradient-Guided Super-Resolution (SwinGSR), a novel dual-branch transformer for image SR. Specifically, Residual Swin Transformer Block (RSTB) is used to extract deep features from the SR branch (SRB) and gradient branch (GB). SwinGSR leverages high-frequency information to prevent edge structure loss and improve image reconstruction with gradient guidance. Furthermore, we propose a Edge-Aware Fusion Module (EAFM), which incorporates several Channel Shuffle Residual Blocks (CSRBs) as its basic blocks and a Multi-Dconv Head Transposed Attention Block (MDTAB) to integrate features from both branches. This approach addresses the Swin Transformer’s limited receptive field and improves feature fusion. The CSRBs enhance generalization by applying channel shuffling. Meanwhile, the MDTAB fully utilizes edge feature information. Extensive experiments demonstrate that SwinGSR outperforms previous methods.</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

Gradient-guided swin transformer with edge-aware fusion for enhanced image super-resolution

  • 期刊论文

  • 发表日期：

    2025年9月16日

查看全文
  • Chentao Qian,
  • Bailin Liu

摘要

Current research on super-resolution (SR) still suffer from undesirable structural distortions, limiting their effectiveness in preserving critical edge and texture details. To address this issue, we propose incorporating gradients as high-frequency information to enhance structural preservation. Building on this idea, we introduce the Swin Transformer-based Gradient-Guided Super-Resolution (SwinGSR), a novel dual-branch transformer for image SR. Specifically, Residual Swin Transformer Block (RSTB) is used to extract deep features from the SR branch (SRB) and gradient branch (GB). SwinGSR leverages high-frequency information to prevent edge structure loss and improve image reconstruction with gradient guidance. Furthermore, we propose a Edge-Aware Fusion Module (EAFM), which incorporates several Channel Shuffle Residual Blocks (CSRBs) as its basic blocks and a Multi-Dconv Head Transposed Attention Block (MDTAB) to integrate features from both branches. This approach addresses the Swin Transformer’s limited receptive field and improves feature fusion. The CSRBs enhance generalization by applying channel shuffling. Meanwhile, the MDTAB fully utilizes edge feature information. Extensive experiments demonstrate that SwinGSR outperforms previous methods.