V-Bridge: Bridging Video Generative Priors to Versatile Few-shot Image Restoration

TL;DR

V-Bridge demonstrates that large-scale video generative models can be repurposed as versatile, data-efficient image restoration tools, achieving competitive results with minimal training data by reinterpreting restoration as a progressive generative process.

Contribution

This work introduces V-Bridge, a novel framework that leverages pretrained video models for multi-task image restoration using limited data, challenging traditional boundaries in low-level vision.

Findings

Pretrained video models can perform competitive image restoration with only 1,000 training samples.

A single model can handle multiple restoration tasks, rivaling specialized architectures.

Video models implicitly learn powerful, transferable restoration priors.

Abstract

Large-scale video generative models are trained on vast and diverse visual data, enabling them to internalize rich structural, semantic, and dynamic priors of the visual world. While these models have demonstrated impressive generative capability, their potential as general-purpose visual learners remains largely untapped. In this work, we introduce V-Bridge, a framework that bridges this latent capacity to versatile few-shot image restoration tasks. We reinterpret image restoration not as a static regression problem, but as a progressive generative process, and leverage video models to simulate the gradual refinement from degraded inputs to high-fidelity outputs. Surprisingly, with only 1,000 multi-task training samples (less than 2% of existing restoration methods), pretrained video models can be induced to perform competitive image restoration, achieving multiple tasks with a single model, rivaling specialized architectures designed explicitly for this purpose. Our findings reveal that video generative models implicitly learn powerful and transferable restoration priors that can be activated with only extremely limited data, challenging the traditional boundary between generative modeling and low-level vision, and opening a new design paradigm for foundation models in visual tasks.