Making Every Step Count: Spatio-Temporal Information Allocation for Imaging Inverse Problems

2026-08-12

AI Summary

This work addresses the suboptimal spatiotemporal resource allocation in existing flow-based solvers for inverse problems, which often leads to insufficient early-stage semantic exploration or limited late-stage detail recovery under a fixed computational budget, compounded by a lack of effective guidance in missing regions. To overcome these limitations, we propose a training-free spatiotemporal information allocation strategy that jointly optimizes time-step scheduling and spatial attention within flow-based inversion. Specifically, Spectral Adaptive Scheduling (SAS) dynamically allocates computational resources across time steps based on spectral content, while Measurement-Prior Attention (MPA) directs spatial information flow toward regions with weaker constraints. This plug-and-play approach achieves consistent performance gains across super-resolution, motion deblurring, and image inpainting tasks, significantly outperforming existing solvers without requiring model retraining or additional inference calls.

Abstract

Flow-based generative models have emerged as powerful image priors for training-free inverse problem solving, capturing coherent semantics and fine-grained structure. Despite these strengths, existing flow-based inverse solvers primarily focus on the design of individual updates, largely overlooking spatio-temporal information allocation under a fixed number of function evaluations (NFEs). Temporally, insufficient early exploration can trap the flow trajectory in an incorrect semantic basin, whereas excessive allocation of NFEs to early stages leaves little budget for late-stage refinement. Spatially, data consistency provides direct constraints only within observed regions, whereas the recovery of missing regions relies mainly on the generative prior. To address these two issues, we introduce two complementary and training-free components, i.e., Spectrum-Adaptive Scheduling (SAS) and Measurement-Prioritized Attention (MPA). For temporal allocation, SAS distributes the available NFEs over flow time according to the degradation spectrum and logSNR geometry, thus better balancing semantic exploration and detail refinement. For spatial propagation, MPA exploits data-prior conflicts to guide information toward weakly constrained regions, thereby enhancing semantic and structural fidelity. Extensive experiments on standard image inverse problems, e.g., super-resolution, motion deblurring, and inpainting, demonstrate that the proposed components can be integrated into existing flow-based inverse solvers in a plug-and-play manner without retraining or additional flow-model evaluations, and can also significantly improve the restoration quality of existing solvers.

Problem

• flow-based generative models
• inverse problems
• spatio-temporal allocation
• function evaluations
• data consistency

Innovation

• flow-based generative models
• spatio-temporal allocation
• training-free inversion
• spectrum-adaptive scheduling
• measurement-prioritized attention