<p>Transformer-based super-resolution (SR) methods have exhibited impressive performance in the field of image super resolution reconstruction. However, these methods face limitations in effectively utilizing prior information and involve a substantial number of irrelevant tokens during computation, which degrade the quality of image reconstruction. To address these issues, this paper proposes the prior-enhanced selectable sparse attention network (PESSANet), a novel framework specifically designed for SR tasks. The hybrid difference convolution block (HDCB) is employed for shallow feature extraction, enabling the capture of rich prior information, thereby enhancing the recovery of image textures. The local-global feature extraction block (LGEFB) is introduced to extract deep features, which consists of multi-head dynamic local self-attention block (MHDLSAB) and top-<i>k</i> selectable sparse attention block (KSSAB). Moreover, KSSAB adopts a more selectable and flexible strategy, which selectively retains the most relevant self-attention values, effectively reducing the impact of irrelevant tokens. The suggested method continuously surpasses current state-of-the-art techniques based on convolutional neural networks (CNNs) and Transformers, producing better results in both image quality and reconstruction accuracy, according to extensive experiments conducted on several benchmark datasets. Extensive experiments are conducted on several widely used benchmark datasets, including Set5, Set14, BSD100, Urban100, and Manga109, under multiple upsampling factors (e.g., <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\times 2\)</EquationSource> </InlineEquation>, <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(\times 3\)</EquationSource> </InlineEquation>, and <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(\times 4\)</EquationSource> </InlineEquation>). Specifically, PESSANet achieved PSNR(dB)/SSIM of 29.00/0.7917, 27.86/0.7464, 27.18/0.8175 on Set14<InlineEquation ID="IEq4"> <EquationSource Format="TEX">\(\times \)</EquationSource> </InlineEquation>4, B100<InlineEquation ID="IEq5"> <EquationSource Format="TEX">\(\times \)</EquationSource> </InlineEquation>4, and Urban100<InlineEquation ID="IEq6"> <EquationSource Format="TEX">\(\times \)</EquationSource> </InlineEquation>4 datasets with low parameter count and fewer floating-point operations.</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

PESSANet: prior-enhanced selectable sparse attention network for image super-resolution

  • Jingjing Liu,
  • Haocan Jiang,
  • Aiying Guo,
  • Jianhua Zhang

摘要

Transformer-based super-resolution (SR) methods have exhibited impressive performance in the field of image super resolution reconstruction. However, these methods face limitations in effectively utilizing prior information and involve a substantial number of irrelevant tokens during computation, which degrade the quality of image reconstruction. To address these issues, this paper proposes the prior-enhanced selectable sparse attention network (PESSANet), a novel framework specifically designed for SR tasks. The hybrid difference convolution block (HDCB) is employed for shallow feature extraction, enabling the capture of rich prior information, thereby enhancing the recovery of image textures. The local-global feature extraction block (LGEFB) is introduced to extract deep features, which consists of multi-head dynamic local self-attention block (MHDLSAB) and top-k selectable sparse attention block (KSSAB). Moreover, KSSAB adopts a more selectable and flexible strategy, which selectively retains the most relevant self-attention values, effectively reducing the impact of irrelevant tokens. The suggested method continuously surpasses current state-of-the-art techniques based on convolutional neural networks (CNNs) and Transformers, producing better results in both image quality and reconstruction accuracy, according to extensive experiments conducted on several benchmark datasets. Extensive experiments are conducted on several widely used benchmark datasets, including Set5, Set14, BSD100, Urban100, and Manga109, under multiple upsampling factors (e.g., \(\times 2\) , \(\times 3\) , and \(\times 4\) ). Specifically, PESSANet achieved PSNR(dB)/SSIM of 29.00/0.7917, 27.86/0.7464, 27.18/0.8175 on Set14 \(\times \) 4, B100 \(\times \) 4, and Urban100 \(\times \) 4 datasets with low parameter count and fewer floating-point operations.