Research map / Image generation / Image restoration
Latent Diffusion Prior Enhanced Deep Unfolding for Snapshot Spectral Compressive Imaging
Zongliang Wu, Ruiying Lu, Ying Fu, Xin Yuan
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 the LDM-generated prior (diffusion) to the unfolding network (ablation) | Ours-3stg · KAIST 10 scenes | PSNR (dB): 37.61 -> 38.40 | Improved |
| feed clean HSI plus measurement (accurate guidance) to the latent encoder (ablation) | Ours-3stg · KAIST 10 scenes | PSNR (dB): 37.63 -> 38.40 | Improved |
| add gradient-correction GAP projection (ablation) | Ours-3stg · KAIST 10 scenes | PSNR (dB): 38.01 -> 38.40 | Improved |
| add spatial flow to the Trident Transformer (ablation) | Ours-3stg · KAIST 10 scenes | PSNR (dB): 37.66 -> 38.40 | Improved |
| add cross prior flow (cross-attention) instead of MLP layers (ablation) | Ours-3stg · KAIST 10 scenes | PSNR (dB): 37.89 -> 38.40 | Improved |
| use ACS-MHSA instead of basic MHSA in cross spectral flow (ablation) | Ours-3stg · KAIST 10 scenes | PSNR (dB) / FLOPs (G): 38.31 -> 38.40 PSNR; 36.80 -> 33.80 FLOPs | Improved |
| increase latent prior length N from 16 to 256 (ablation) | Ours-3stg · KAIST 10 scenes | PSNR (dB): 38.40 -> 38.41 (FLOPs 33.80 -> 36.99) | No clear effect |
| decrease latent prior length N from 16 to 4 (ablation) | Ours-3stg · KAIST 10 scenes | PSNR (dB): 38.40 -> 37.99 | Hurt |
| use proposed 9-stage model instead of RDLUF-MixS2-9stg | deep unfolding CASSI · KAIST 10 scenes | FLOPs (G): 231.09 -> 88.68 (PSNR 39.57 -> 40.09) | Improved |
| use proposed 9-stage model with diffusion instead of PADUT-12stg | deep unfolding CASSI · KAIST 10 scenes | Inference time (ms): 749.94 -> 1096.58 | Hurt |
Closest papers by meaning
- Dual Prior Unfolding for Snapshot Compressive Imaging (CVPR 2024 · similarity 0.89)
- DiffSCI: Zero-Shot Snapshot Compressive Imaging via Iterative Spectral Diffusion Model (CVPR 2024 · similarity 0.89)
- Using Powerful Prior Knowledge of Diffusion Model in Deep Unfolding Networks for Image Compressive Sensing (CVPR 2025 · similarity 0.88)
- Untrained Neural Nets for Snapshot Compressive Imaging: Theory and Algorithms (NeurIPS 2024 · similarity 0.87)
- Detail Matters: Mamba-Inspired Joint Unfolding Network for Snapshot Spectral Compressive Imaging (AAAI 2025 · similarity 0.87)
- Accelerated Diffusion via High-Low Frequency Decomposition for Pan-Sharpening (AAAI 2025 · similarity 0.87)
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.
