Research map / Image generation / Image restoration
Intrinsic Image Decomposition Using Paradigms
David Alexander Forsyth, Jason Rock
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 adversarial smoothing (ablation) | BBA paradigm intrinsic decomposition · IIW simulated test sets | WHDR: Without adversarial smoothing (NoSmo), performance is comparable to Retinex | Improved |
| add averaging over more tiles and scales (ablation) | Base paradigm method · IIW simulated test sets | WHDR: N_t=7, N_sigma=3 (Base) -> N_t=15, N_sigma=5 (BBA) | Improved |
| add discrete image averaging (ablation) | BBAP · IIW simulated test sets | WHDR: BBAF always reports a very much better WHDR than BBA | Improved |
| use discrete image averaging instead of discrete tile averaging (ablation) | BBAP · IIW simulated test sets | WHDR: BBAF clearly better than BBAT | Improved |
| add postprocessing (ablation) | BBA · IIW test (held-out threshold) | WHDR: BBAP appears worse than BBA on standard test; BBAP almost always smaller WHDR on simulated sets | Mixed |
| use CGIntrinsics tiles instead of paradigm images (ablation) | BBAF · IIW simulated test sets | WHDR: outperformed by Retinex | Hurt |
| vary paradigm details (fragment sizes) (ablation) | BBAF · IIW simulated test sets | WHDR | No clear effect |
| use a dark shading paradigm (ablation) | BBAF · IIW simulated test sets | WHDR | Hurt |
| vary discriminator patch scale from 22x22 to 10x10, 29x29, 48x48 or 128x128 (ablation) | BBAF · IIW simulated test sets | WHDR: 22 x 22 best; SD 10 x 10, ID 29 x 29, MD 48 x 48, BD 128 x 128 | Mixed |
| add extra heads for thin and thick dark bars | BBAF · IIW test | WHDR | Hurt |
| choose models by single standard test WHDR vs many simulated test sets | evaluation protocol · IIW simulated test sets | WHDR std: standard deviation is 0.3% for the base method | Mixed |
Closest papers by meaning
- A Survey on Intrinsic Images: Delving Deep into Lambert and Beyond (IJCV 2022 · similarity 0.90)
- Learning Intrinsic Image Decomposition From Watching the World (CVPR 2018 · similarity 0.89)
- Reflectance Adaptive Filtering Improves Intrinsic Image Estimation (CVPR 2017 · similarity 0.89)
- ReasonX: MLLM-Guided Intrinsic Image Decomposition (CVPR 2026 · similarity 0.87)
- Intrinsic Image Decomposition for Robust Self-supervised Monocular Depth Estimation on Reflective Surfaces (AAAI 2025 · similarity 0.86)
- Latent Intrinsics Emerge from Training to Relight (NeurIPS 2024 · 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.
