Zero-Shot Video Restoration and Enhancement with Assistance of Video Diffusion Models

TL;DR

This paper introduces a novel, training-free framework that leverages video diffusion models to enhance zero-shot video restoration and enhancement, significantly reducing temporal flickering and improving consistency.

Contribution

It presents the first framework combining image-based methods with video diffusion models for zero-shot video restoration, introducing fusion strategies and post-processing for better temporal coherence.

Findings

Outperforms existing methods in temporal consistency.

Effective fusion strategies improve restoration quality.

Training-free approach applicable to various diffusion models.

Abstract

Although diffusion-based zero-shot image restoration and enhancement methods have achieved great success, applying them to video restoration or enhancement will lead to severe temporal flickering. In this paper, we propose the first framework that utilizes the rapidly-developed video diffusion model to assist the image-based method in maintaining more temporal consistency for zero-shot video restoration and enhancement. We propose homologous latents fusion, heterogenous latents fusion, and a COT-based fusion ratio strategy to utilize both homologous and heterogenous text-to-video diffusion models to complement the image method. Moreover, we propose temporal-strengthening post-processing to utilize the image-to-video diffusion model to further improve temporal consistency. Our method is training-free and can be applied to any diffusion-based image restoration and enhancement methods. Experimental results demonstrate the superiority of the proposed method.