Reproduce “A Systematic Survey of Deep Learning-based Single-Image Super-Resolution” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Juncheng Li, Zehua Pei, Wenjie Li, Guangwei Gao, Longguang Wang, Yingqian Wang, Tieyong Zeng
This survey provides a systematic review of over 100 deep learning-based Single-Image Super-Resolution methods, categorizing them by their specific design targets rather than by supervision type. It synthesizes architectural advancements and evaluates performance across various models, identifying critical challenges and future research directions for more practical and versatile SR solutions.
Key insights
- Deep learning has transformed SISR by enabling efficient end-to-end solutions, with significant progress in network design (e.g., residual, dense, Transformer architectures) and perceptual quality (e.g., GANs, diffusion models).
- Leveraging additional information (internal statistics, priors, reference images) and adapting models for real-world degradations and specific domains are crucial for robust performance.
- Effective SISR model design prioritizes architectural efficiency over merely increasing parameter count, and performance often scales with the quality and size of training datasets.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.