Bridging Restoration and Generation Manifolds in One-Step Diffusion for Real-World Super-Resolution

2026-04-27

AI Summary

Existing single-step diffusion-based super-resolution methods struggle to balance perceptual quality and fidelity due to fixed timestep initialization, mismatched trajectory distributions, and sensitivity to stochastic modulation. This work proposes IDaS-SR, a framework that jointly models deterministic restoration and stochastic generative manifolds to achieve high-quality real-world super-resolution within a single inference step. The approach introduces two key innovations: the MINE module, which adaptively predicts degradation-aware noise levels and timesteps, and the CHARIOT mechanism, which flexibly modulates the perception-distortion trade-off through continuous trajectory rescheduling and noise interpolation while preserving structural priors. Built upon a distilled architecture from pretrained diffusion models, IDaS-SR significantly outperforms current methods across multiple real-world super-resolution benchmarks, demonstrating superior efficiency, visual quality, and structural fidelity.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Sampling/Simulation-based Search
• Machine Learning: Deep Generative Models & Autoencoders

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Pretrained diffusion models have revolutionized real-world image super-resolution (Real-ISR) but suffer from computational bottlenecks due to iterative sampling. Recent single-step distillation accelerates inference but faces a stark perception-distortion trade-off due to rigid timestep initialization, distributional trajectory mismatches, and fragile stochastic modulation. To address this, we present Adaptive Inversion and Degradation-aware Sampling for Real-ISR (IDaS-SR), a one-step framework bridging the deterministic restoration and stochastic generation manifolds. At its core, the Manifold Inversion Noise Estimator (MINE) resolves these initialization and trajectory mismatches by predicting a severity-aware timestep and inversion noise, precisely anchoring low-quality latents onto the diffusion trajectory. Furthermore, to mitigate fragile stochastic modulation, we propose CHARIOT, a continuous generative steering mechanism. By rescheduling trajectories and interpolating noise, it enables explicit navigation of the perception-distortion boundary without compromising structural priors. Extensive experiments demonstrate that IDaS-SR outperforms state-of-the-art methods, seamlessly transitioning from a rigorous structural restorer to a sophisticated texture hallucinator in a single inference step.

Problem

• super-resolution
• diffusion models
• perception-distortion trade-off
• one-step generation
• real-world image restoration

Innovation

• one-step diffusion
• manifold bridging
• adaptive inversion
• perception-distortion trade-off
• degradation-aware sampling