Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Single Image Super-Resolution Based on Global-Local Information Synergy

Nianzu Qiao, Lamei Di, Changyin Sun

TL;DR

This work tackles the trade-off between high-accuracy single-image super-resolution and computational efficiency. It introduces a Global-Local Information Synergy framework built from a Basic Block Module (SCAM and CFC with a feature-map dot-product fusion) and a Global-Local Information Extraction Module, followed by PixelShuffle upsampling. A combined loss with pixel-wise MAE and FFT-based frequency guidance (γ = 0.05) drives reconstruction of both structure and texture: $L = ||I_{SR}-I_{HR}|| + γ || φ(I_{SR})-φ(I_{HR}) ||$. Empirically, the method achieves competitive or superior performance on standard benchmarks while substantially reducing parameters and FLOPs, indicating strong practical potential for efficient SR deployments.

Abstract

Although several image super-resolution solutions exist, they still face many challenges. CNN-based algorithms, despite the reduction in computational complexity, still need to improve their accuracy. While Transformer-based algorithms have higher accuracy, their ultra-high computational complexity makes them difficult to be accepted in practical applications. To overcome the existing challenges, a novel super-resolution reconstruction algorithm is proposed in this paper. The algorithm achieves a significant increase in accuracy through a unique design while maintaining a low complexity. The core of the algorithm lies in its cleverly designed Global-Local Information Extraction Module and Basic Block Module. By combining global and local information, the Global-Local Information Extraction Module aims to understand the image content more comprehensively so as to recover the global structure and local details in the image more accurately, which provides rich information support for the subsequent reconstruction process. Experimental results show that the comprehensive performance of the algorithm proposed in this paper is optimal, providing an efficient and practical new solution in the field of super-resolution reconstruction.

Single Image Super-Resolution Based on Global-Local Information Synergy

TL;DR

This work tackles the trade-off between high-accuracy single-image super-resolution and computational efficiency. It introduces a Global-Local Information Synergy framework built from a Basic Block Module (SCAM and CFC with a feature-map dot-product fusion) and a Global-Local Information Extraction Module, followed by PixelShuffle upsampling. A combined loss with pixel-wise MAE and FFT-based frequency guidance (γ = 0.05) drives reconstruction of both structure and texture: . Empirically, the method achieves competitive or superior performance on standard benchmarks while substantially reducing parameters and FLOPs, indicating strong practical potential for efficient SR deployments.

Abstract

Although several image super-resolution solutions exist, they still face many challenges. CNN-based algorithms, despite the reduction in computational complexity, still need to improve their accuracy. While Transformer-based algorithms have higher accuracy, their ultra-high computational complexity makes them difficult to be accepted in practical applications. To overcome the existing challenges, a novel super-resolution reconstruction algorithm is proposed in this paper. The algorithm achieves a significant increase in accuracy through a unique design while maintaining a low complexity. The core of the algorithm lies in its cleverly designed Global-Local Information Extraction Module and Basic Block Module. By combining global and local information, the Global-Local Information Extraction Module aims to understand the image content more comprehensively so as to recover the global structure and local details in the image more accurately, which provides rich information support for the subsequent reconstruction process. Experimental results show that the comprehensive performance of the algorithm proposed in this paper is optimal, providing an efficient and practical new solution in the field of super-resolution reconstruction.
Paper Structure (18 sections, 5 equations, 5 figures, 4 tables)

This paper contains 18 sections, 5 equations, 5 figures, 4 tables.

Table of Contents

  1. Introduction
  2. The recommended approach
  3. Basic Block Module
  4. SCAM Module
  5. CFC Module
  6. Global-Local Information Extraction Module
  7. Loss Function
  8. Experiments
  9. Experimental Settings
  10. Datasets
  11. Assessment Indicators
  12. Performance Comparison
  13. Quantitative Results
  14. Qualitative Results
  15. Ablation Analysis
  16. ...and 3 more sections

Figures (5)

  • Figure 1: Schematic structure of deep learning algorithm for image super-resolution reconstruction based on global-local information.
  • Figure 2: Diagram of Basic Block Module.
  • Figure 3: Diagram of global-local information extraction module.
  • Figure 4: Qualitative comparison results of different algorithms on the Urban100 dataset, the first image is the $\times 2$-scale result, the second image is the $\times 3$-scale result, and the third image is the $\times 4$-scale result. (a) LR images, (b) EDSR results, (c) RCAN results, (d) HAN results, (e) SAFMN results, (f) results from this chapter, (g) SRFormer results, (h) HR images.
  • Figure 5: Qualitative comparison results of different components on Urban100 dataset ($\times 3$ scales). (a) LR image, (b) -w/o CFC results, (c) -w/o SCAM results, (d) -w/o Global-Local Information Extraction Module results, (e) results with all components included, (f) HR image