FlowSteer: Conditioning Flow Field for Consistent Image Restoration

TL;DR

FlowSteer addresses the mismatch between flow-based image generation priors and pixel-level fidelity required in restoration. By introducing a simple, training-free scheduler that explicitly gates forward-model conditioning during the reconstruction path, FlowSteer reconciles measurement fidelity with rich generative priors in zero-shot fashion. The method, demonstrated on colorization, super-resolution, deblurring, and denoising, leverages a pre-trained flow model (Flux FLUX2024) with implicit attentional guidance augmented by a sparse fidelity-update schedule, yielding higher PSNR/LPIPS and preserved identity. The results highlight FlowSteer’s practicality as a universal, post-hoc conditioning mechanism for off-the-shelf flow models, with clear guidelines for schedule design and notable ablations clarifying the role of timing and artifacts.

Abstract

Flow-based text-to-image (T2I) models excel at prompt-driven image generation, but falter on Image Restoration (IR), often "drifting away" from being faithful to the measurement. Prior work mitigate this drift with data-specific flows or task-specific adapters that are computationally heavy and not scalable across tasks. This raises the question "Can't we efficiently manipulate the existing generative capabilities of a flow model?" To this end, we introduce FlowSteer (FS), an operator-aware conditioning scheme that injects measurement priors along the sampling path,coupling a frozed flow's implicit guidance with explicit measurement constraints. Across super-resolution, deblurring, denoising, and colorization, FS improves measurement consistency and identity preservation in a strictly zero-shot setting-no retrained models, no adapters. We show how the nature of flow models and their sensitivities to noise inform the design of such a scheduler. FlowSteer, although simple, achieves a higher fidelity of reconstructed images, while leveraging the rich generative priors of flow models.

Figures (14)

• Figure 1: Image inversion(left) and reconstruction(right) paths of a flow model. A pre-trained flow trajectory $v_{\theta}$ is conditioned implicitly by language prompts ($C_1 , C_2)$ and feature sharing between paths. FlowSteer introduces an explicit conditioning schedule (center) to steer the reconstruction path toward high pixel-level fidelity without retraining the model.
• Figure 2: Effect of varying classifier-free guidance $\gamma$ and feature-sharing steps $\zeta$ on restoration quality. Increasing $\gamma$ enhances color saturation but reduces fidelity to the measurement $y$, while increasing $\zeta$ improves pixel-level consistency without recovering true color.
• Figure 3: Linear projections of flow models exacerbate the errors from non-ideal velocities. We recommend to avoid projection operations to estimate $\mathbf{\widehat{x}}_{0|t_i}$ when $t_i \approx 1$.
• Figure 4: Colorizing with different $\{ \lambda_i\}$ schedules on our baseline RF model. No explicit conditioning creates hallucinations and loses identity. Having a constant conditioning with $\lambda = 1$ introduces undesired noise.
• Figure 5: Qualitative comparison of flow-based methods. Columns 2-4 are image restoration models, and column 5 is an image editing model. The rows show four degradations: tasks focused on information reconstruction (colorization and $4\times$ super-resolution) and tasks focused on corruption removal (denoising and deblurring). SteerFlow removes degradations, and details are generated without losing the identity of the subject.