Research map / Image generation / Super-resolution
Improving Scene Text Image Super-resolution via Dual Prior Modulation Network
Shipeng Zhu, Zuoyan Zhao, Pengfei Fang, Hui Xue
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add DPMN to TATT | TATT · TextZoom | PSNR: 21.21 -> 21.49 | Improved |
| Add DPMN to TATT | TATT · TextZoom | SSIM: 0.7825 -> 0.7925 | Improved |
| Add DPMN to TATT | TATT · TextZoom | Accuracy: 58.90 -> 59.22 | Improved |
| Use complementary mask and graphic-recognition priors in separate branches rather than a single branch (ablation) | TATT + DPMN · TextZoom | PSNR: 21.01 -> 21.49 | Improved |
| Use complementary mask and graphic-recognition priors in separate branches rather than a single branch (ablation) | TATT + DPMN · TextZoom | Accuracy: 63.67 -> 63.89 | Improved |
| Add the graphic-recognition-result prior alongside the mask prior (ablation) | TATT + DPMN · TextZoom | PSNR: 21.12 -> 21.01 | No clear effect |
| Use HR-derived mask and GRR priors (ablation) | TATT + DPMN · TextZoom | PSNR: 21.49 -> 23.60 | Improved |
| Train DPMN without a pretrained STISR backbone (ablation) | DPMN · TextZoom Average, ASTER | Accuracy: 63.89 -> 49.42 | Hurt |
| Freeze the pretrained PSN while training DPMN instead of fine-tuning the PSN (ablation) | TATT + DPMN · TextZoom Average, ASTER | Accuracy: 63.30 -> 63.89 | Improved |
| Use 3 PGRMs per branch instead of 1, 2, 4, or 5 (ablation) | TATT + DPMN · TextZoom Average, ASTER | Accuracy: 63.71, 63.57, 63.60, 63.60 -> 63.89 | Improved |
| Replace dynamic windows with fixed window size 2, 4, or 8 (ablation) | TATT + DPMN · TextZoom Average, ASTER | Accuracy: 63.59, 63.55, 63.64 -> 63.89 | Improved |
| Remove channel attention from the encoder-decoder CMM (ablation) | TATT + DPMN · TextZoom Average, ASTER | Accuracy: 63.55 -> 63.89 | Improved |
Closest papers by meaning
- Text-Aware Real-World Image Super-Resolution via Diffusion Model with Joint Segmentation Decoders (NeurIPS 2025 · similarity 0.89)
- The Power of Context: How Multimodality Improves Image Super-Resolution (CVPR 2025 · similarity 0.88)
- SRSR: Enhancing Semantic Accuracy in Real-World Image Super-Resolution with Spatially Re-Focused Text-Conditioning (NeurIPS 2025 · similarity 0.87)
- PRISM: Prior Rectification and Uncertainty-Aware Structure Modeling for Diffusion-Based Text Image Super-Resolution (NeurIPS 2026 · similarity 0.87)
- GlyphSR: A Simple Glyph-Aware Framework for Scene Text Image Super-Resolution (AAAI 2025 · similarity 0.86)
- SeeSR: Towards Semantics-Aware Real-World Image Super-Resolution (CVPR 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
