Research map / Image generation / Super-resolution
MR Image Super-Resolution With Squeeze and Excitation Reasoning Attention Network
Yulun Zhang, Kai Li, Kunpeng Li, Yun Fu
Check 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 |
|---|---|---|---|
| Use SERAN instead of CSN | MR image super-resolution · PD, T1, and T2 at ×2, ×3, and ×4 | PSNR (dB): PD: 41.28 -> 41.53 (×2), 35.87 -> 36.17 (×3), 33.40 -> 33.77 (×4); T1: 38.27 -> 38.66 (×2), 33.53 -> 34.08 (×3), 31.23 - | Improved |
| Apply geometric self-ensemble | SERAN · PD, T1, and T2 at ×2, ×3, and ×4 | PSNR (dB): PD: 41.53 -> 41.66 (×2), 36.17 -> 36.34 (×3), 33.77 -> 33.97 (×4); T1: 38.66 -> 38.74 (×2), 34.08 -> 34.18 (×3), 31.89 - | Improved |
| Add SEAB without residual scaling (ablation) | Residual network (40 residual blocks) · PD, T1, and T2 ×2 | PSNR (dB): PD: 40.68 -> 41.31; T1: 37.82 -> 38.41; T2: 38.88 -> 39.80 | Improved |
| Add residual scaling to SEAB (ablation) | Residual network with SEAB · PD, T1, and T2 ×2 | PSNR (dB): PD: 41.31 -> 41.43; T1: 38.41 -> 38.56; T2: 39.80 -> 40.07 | Improved |
| Add primitive relationship reasoning with GCN (ablation) | Residual network with SEAB and residual scaling · PD, T1, and T2 ×2 | PSNR (dB): PD: 41.43 -> 41.53; T1: 38.56 -> 38.66; T2: 40.07 -> 40.18 | Improved |
| Place the single SERAB in a higher-level residual block rather than a lower-level block (ablation) | Residual network · PD, T1, and T2 ×2 | PSNR (dB): PD: 41.38 -> 41.41 -> 41.53; T1: 38.32 -> 38.43 -> 38.66; T2: 39.88 -> 39.96 -> 40.18 | Improved |
| Add more SERAB blocks at low-, middle-, and high-level positions (ablation) | Residual network · PD, T1, and T2 ×2 | PSNR (dB): With all three positions: PD 41.62; T1 38.76; T2 40.29. Compared with a single high-level block: PD 41.53; T1 38.66; T2 | Improved |
Closest papers by meaning
- Second-Order Attention Network for Single Image Super-Resolution (CVPR 2019 · similarity 0.88)
- Context Reasoning Attention Network for Image Super-Resolution (ICCV 2021 · similarity 0.88)
- Rethinking Multi-Contrast MRI Super-Resolution: Rectangle-Window Cross-Attention Transformer and Arbitrary-Scale Upsampling (ICCV 2023 · similarity 0.87)
- MFTN: Multi-Level Feature Transfer Network Based on MRI-Transformer for MR Image Super-resolution (AAAI 2024 · similarity 0.86)
- Boosting ViT-based MRI Reconstruction from the Perspectives of Frequency Modulation, Spatial Purification, and Scale Diversification (AAAI 2025 · similarity 0.85)
- A Unified Model for Compressed Sensing MRI Across Undersampling Patterns (CVPR 2025 · similarity 0.85)
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.
