Adaptive Domain Shift in Diffusion Models for Cross-Modality Image Translation

2026-01-26

AI Summary

Existing cross-modal image translation methods rely on global linear domain shifts, which often lead to sampling trajectories deviating from the data manifold, causing semantic drift and inefficiency. This work proposes an adaptive domain shift mechanism that dynamically embeds spatially varying mixture fields during continuous-time diffusion and introduces an explicit target-consistent restoration term to correct such drift. By reframing the model’s role from global alignment to local residual correction, the approach admits a closed-form solution, preserves boundary consistency, and significantly enhances structural fidelity and semantic coherence while reducing the required number of denoising steps. Its effectiveness is validated across diverse applications, including medical imaging, remote sensing, and electroluminescence semantic mapping.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Accountability, Transparency, and Ethics for personalization

Abstract

Cross-modal image translation remains brittle and inefficient. Standard diffusion approaches often rely on a single, global linear transfer between domains. We find that this shortcut forces the sampler to traverse off-manifold, high-cost regions, inflating the correction burden and inviting semantic drift. We refer to this shared failure mode as fixed-schedule domain transfer. In this paper, we embed domain-shift dynamics directly into the generative process. Our model predicts a spatially varying mixing field at every reverse step and injects an explicit, target-consistent restoration term into the drift. This in-step guidance keeps large updates on-manifold and shifts the model's role from global alignment to local residual correction. We provide a continuous-time formulation with an exact solution form and derive a practical first-order sampler that preserves marginal consistency. Empirically, across translation tasks in medical imaging, remote sensing, and electroluminescence semantic mapping, our framework improves structural fidelity and semantic consistency while converging in fewer denoising steps.

Problem

• cross-modality image translation
• domain shift
• diffusion models
• semantic drift
• fixed-schedule domain transfer

Innovation

• adaptive domain shift
• diffusion models
• cross-modality image translation
• spatially varying mixing field
• on-manifold guidance