LaverNet: Lightweight All-in-one Video Restoration via Selective Propagation

2025-12-18

AI Summary

Existing video restoration methods struggle with time-varying degradations in multi-task settings, where degradation interference impedes accurate temporal modeling, and large-scale masked modeling approaches suffer from inherent architectural limitations. Method: This paper proposes LaverNet—a lightweight, unified convolutional model (362K parameters)—featuring a novel selective cross-frame propagation mechanism. By integrating degradation-aware feature gating and spatiotemporal decoupling, LaverNet explicitly disentangles degradation features from semantic content features, enabling degradation-agnostic, high-fidelity temporal modeling. It supports end-to-end joint training across multiple degradation types. Contribution/Results: On multiple benchmarks, LaverNet achieves state-of-the-art (SOTA) or comparable performance to leading large models while using less than 1% of their parameters. It significantly improves inference speed and memory efficiency, demonstrating for the first time the feasibility and superiority of lightweight, full-task video restoration.

Technology Category

• Computer Vision: Large Vision Models
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Recent studies have explored all-in-one video restoration, which handles multiple degradations with a unified model. However, these approaches still face two challenges when dealing with time-varying degradations. First, the degradation can dominate temporal modeling, confusing the model to focus on artifacts rather than the video content. Second, current methods typically rely on large models to handle all-in-one restoration, concealing those underlying difficulties. To address these challenges, we propose a lightweight all-in-one video restoration network, LaverNet, with only 362K parameters. To mitigate the impact of degradations on temporal modeling, we introduce a novel propagation mechanism that selectively transmits only degradation-agnostic features across frames. Through LaverNet, we demonstrate that strong all-in-one restoration can be achieved with a compact network. Despite its small size, less than 1% of the parameters of existing models, LaverNet achieves comparable, even superior performance across benchmarks.

Problem

• Addresses time-varying degradation challenges in video restoration
• Proposes lightweight network to reduce model size and complexity
• Introduces selective propagation to focus on content over artifacts

Innovation

• Lightweight network with only 362K parameters
• Selective propagation of degradation-agnostic features
• Compact model achieving comparable or superior performance