Research map / Image generation / Editing and style transfer
BlindHarmony: "Blind" Harmonization for MR Images via Flow Model
Hwihun Jeong, Heejoon Byun, Dong Un Kang, Jongho Lee
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 |
|---|---|---|---|
| apply BlindHarmony to simulated unseen domains versus unharmonized source | BlindHarmony · OASIS3 simulated domains | PSNR: 21.9 dB -> 28.6 dB (average) | Improved |
| apply BlindHarmony to real scanner domains versus unharmonized source | BlindHarmony · OASIS3 real domains | PSNR: 21.5 dB -> 22.2 dB (average) | Improved |
| apply per-domain U-net or CycleGAN to a mismatched simulated domain | U-net / CycleGAN · OASIS3 simulated Domain-Exp | PSNR: 22.6 -> 16.8 (U-net Log on Exp) | Hurt |
| use per-domain trained U-net/CycleGAN instead of single BlindHarmony on matched simulated domain | U-net / CycleGAN · OASIS3 simulated Domain-Exp | PSNR: 29.6 -> 65.6 (U-net Exp) | Improved |
| apply BlindHarmony on downstream white matter segmentation | BlindHarmony · OASIS3 real Domain 2 | IoU: 0.845 -> 0.878 | Improved |
| apply BlindHarmony on segmentation for domains already close to target | BlindHarmony · OASIS3 real Domain 3 and 4 | IoU: 0.947 -> 0.947 (Domain 3) | No clear effect |
| use histogram matching, SSIMH, U-net or CycleGAN on segmentation input | baseline harmonizers · OASIS3 real Domain 1 | IoU: 0.912 -> 0.863 (HM), 0.306 (CycleGAN) | Hurt |
| apply BlindHarmony with default hyperparameters on extreme Gamma 1.5 simulated domain | BlindHarmony · OASIS3 simulated Gamma 1.5 | PSNR: BlindHarmony 25.7 vs HM 26.5 | Hurt |
| lower beta1 from 1000 to 500 (ablation) | BlindHarmony · OASIS3 simulated Gamma 1.5 | PSNR: 25.7 -> 26.8 | Improved |
Closest papers by meaning
- Patch2Self2: Self-supervised Denoising on Coresets via Matrix Sketching (CVPR 2024 · similarity 0.83)
- S2S2: Semantic Stacking for Robust Semantic Segmentation in Medical Imaging (AAAI 2025 · similarity 0.83)
- Motion Artifact Removal in Pixel-Frequency Domain via Alternate Masks and Diffusion Model (AAAI 2025 · similarity 0.83)
- Perturbating, Tuning, and Collaborating: Harnessing Vision Foundation Models for Single Domain Generalization on Medical Imaging (AAAI 2025 · similarity 0.82)
- Boosting ViT-based MRI Reconstruction from the Perspectives of Frequency Modulation, Spatial Purification, and Scale Diversification (AAAI 2025 · similarity 0.82)
- A Unified Degradation-Robust Approach to SSL and UDA for 3D Medical Images (AAAI 2025 · similarity 0.82)
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.
