Research map / Image generation / Image restoration
Reflectance Adaptive Filtering Improves Intrinsic Image Estimation
Thomas Nestmeyer, Peter V. Gehler
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 |
|---|---|---|---|
| Train a dense reflectance CNN directly with WHDR-Hinge loss rather than relying on pairwise classification and a globalization step | Direct CNN on IIW · IIW test split | WHDR: 19.5 | Improved |
| Apply bilateral filtering with flat guidance to the CNN reflectance | Direct CNN · IIW test split | WHDR: 19.49 -> 18.1 | Improved |
| Apply guided filtering with flat guidance to the CNN reflectance | Direct CNN · IIW test split | WHDR: 19.49 -> 17.7 | Improved |
| Apply guided filtering with flat guidance | Zoran et al. 2015 · IIW test split | WHDR: 17.85 -> 16.38 | Improved |
| Apply the guided filter repeatedly, up to three applications, with flat guidance | Zoran et al. 2015 · IIW test split of Zoran et al. 2015 | WHDR: down to 15.78 after three applications | Improved |
| Filter CNN reflectance using the input image as bilateral guidance | Direct CNN · IIW test split | WHDR: 19.5 -> 18.9 | Improved |
| Filter CNN reflectance using the input image as guided-filter guidance | Direct CNN · IIW test split | WHDR: 19.5 -> 19.2 | Improved |
| Replace the CNN with its L1-flattened guidance image as the reflectance prediction (ablation) | Direct CNN · IIW test split | WHDR: 19.5 -> 20.9 | Hurt |
| Increase convolution kernel size beyond 1x1 (ablation) | Direct CNN · IIW validation set | WHDR | No clear effect |
| Use larger kernel sizes (ablation) | Direct CNN · IIW validation set | WHDR | Hurt |
| Use alternative layouts without skip connections or a U-net, PReLU activations, or dilated convolutions (ablation) | Direct CNN · IIW validation set | WHDR | No clear effect |
| Augment pairwise labels with transitive-closure comparisons (ablation) | Direct CNN training · IIW test set evaluated on original comparisons | WHDR | No clear effect |
Closest papers by meaning
- Unsupervised Learning for Intrinsic Image Decomposition From a Single Image (CVPR 2020 · similarity 0.92)
- Revisiting Deep Intrinsic Image Decompositions (CVPR 2018 · similarity 0.91)
- Unsupervised Intrinsic Image Decomposition Using Internal Self-Similarity Cues (TPAMI 2021 · similarity 0.91)
- A Survey on Intrinsic Images: Delving Deep into Lambert and Beyond (IJCV 2022 · similarity 0.91)
- Learning Intrinsic Image Decomposition From Watching the World (CVPR 2018 · similarity 0.89)
- When Shadow Removal Meets Intrinsic Image Decomposition: A Joint Learning Framework Using Unpaired Data (AAAI 2025 · similarity 0.88)
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.
