UniH^3: Unifying Hierarchical Homogeneity and Heterogeneity for All-in-One Medical Image Restoration
UniH3 uses shared anatomical structure as a training signal, not just task-specific degradation cues.
The paper proposes an all-in-one medical image restoration framework for multiple modalities and degradation types. Its H2M module learns homogeneity priors from high-quality images during training and retrieves relevant ones for each input. Its H2B component is meant to reduce conflicts between and within tasks during optimization. The authors report state-of-the-art results on MedIR-2D-500K and MedIR-3D-3K, and note that the paper has been accepted by ECCV 2026. HF Daily Papers' note
The paper proposes an all-in-one medical image restoration framework for multiple modalities and degradation types. Its H2M module learns homogeneity priors from high-quality images during training and retrieves relevant ones for each input. Its H2B component is meant to reduce conflicts between and within tasks during optimization. The authors report state-of-the-art results on MedIR-2D-500K and MedIR-3D-3K, and note that the paper has been accepted by ECCV 2026. HF Daily Papers' note
score 4