Research map / Image generation / Image restoration
Multi-Block-Attention-based Color Constancy
Oguzhan Ulucan, Diclehan Ulucan, Marc Ebner
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 statistics branch (ablation) | HybridCC · Avg of Gehler-Shi, NUS-8, Cube+ (ablation) | mean angular error: 2.33 -> 1.97 | Improved |
| add backbone branch (ablation) | HybridCC · Avg of Gehler-Shi, NUS-8, Cube+ (ablation) | mean angular error: 2.34 -> 1.97 | Improved |
| add statistics branch for cross-camera generalization (ablation) | HybridCC · NUS-8 (cross-sensor) | mean angular error: 2.01 -> 1.58 | Improved |
| add brightness weighting to soft-max descriptor (ablation) | HybridCC · Avg of Gehler-Shi, NUS-8, Cube+ (ablation) | mean angular error: 2.10 -> 1.97 | Improved |
| initialize tau at 0.02 instead of 0.5 (ablation) | HybridCC · Avg of Gehler-Shi, NUS-8, Cube+ (ablation) | mean angular error: 2.11 -> 1.97 | Improved |
| add learned saliency (ablation) | HybridCC · Avg of Gehler-Shi, NUS-8, Cube+ (ablation) | mean angular error: 2.06 -> 1.97 | Improved |
| add multi-head self-attention across blocks (ablation) | HybridCC · Avg of Gehler-Shi, NUS-8, Cube+ (ablation) | mean angular error: 2.07 -> 1.97 | Improved |
| add spatial mixing (ablation) | HybridCC · Avg of Gehler-Shi, NUS-8, Cube+ (ablation) | mean angular error: 2.09 -> 1.97 | Improved |
| increase spatial mixing layers from 2 to 3 (ablation) | HybridCC · Avg of Gehler-Shi, NUS-8, Cube+ (ablation) | mean angular error: 1.97 -> 2.08 | No clear effect |
| add sparsity loss (ablation) | HybridCC · Avg of Gehler-Shi, NUS-8, Cube+ (ablation) | mean angular error: 2.14 -> 1.97 | Improved |
| use learned block weights instead of equal weights (ablation) | HybridCC · Avg of Gehler-Shi, NUS-8, Cube+ (ablation) | mean angular error: 2.17 -> 1.97 | Improved |
| use 32x32 blocks instead of 96x96 (ablation) | HybridCC · Avg of Gehler-Shi, NUS-8, Cube+ (ablation) | mean angular error: 2.23 -> 1.97 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CCMNet — CCMNet: Leveraging Calibrated Color Correction Matrices for Cross-Camera Color Constancy (ICCV 2025) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- A Multi-Hypothesis Approach to Color Constancy (CVPR 2020 · similarity 0.90)
- FC4: Fully Convolutional Color Constancy With Confidence-Weighted Pooling (CVPR 2017 · similarity 0.89)
- Multi-Domain Learning for Accurate and Few-Shot Color Constancy (CVPR 2020 · similarity 0.88)
- GCC: Generative Color Constancy via Diffusing a Color Checker (CVPR 2025 · similarity 0.88)
- White-Balance First, Adjust Later: Cross-Camera Color Constancy via Vision-Language Evaluation (CVPR 2026 · similarity 0.85)
- Attentive Illumination Decomposition Model for Multi-Illuminant White Balancing (CVPR 2024 · similarity 0.84)
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.
