No Labels, No Look-Ahead: Unsupervised Online Video Stabilization with Classical Priors

TL;DR

This paper introduces an unsupervised online video stabilization method that leverages classical priors and a multithreaded buffer, effectively addressing data scarcity, controllability, and hardware efficiency issues, and demonstrating superior performance on UAV aerial videos.

Contribution

The paper presents a novel unsupervised, classical-prior-based online stabilization framework with a new UAV aerial video dataset, outperforming existing methods in challenging scenarios.

Findings

Outperforms state-of-the-art online stabilizers in quantitative metrics

Achieves performance comparable to offline stabilization methods

Introduces a new UAV aerial video dataset for stabilization evaluation

Abstract

We propose a new unsupervised framework for online video stabilization. Unlike methods based on deep learning that require paired stable and unstable datasets, our approach instantiates the classical stabilization pipeline with three stages and incorporates a multithreaded buffering mechanism. This design addresses three longstanding challenges in end-to-end learning: limited data, poor controllability, and inefficiency on hardware with constrained resources. Existing benchmarks focus mainly on handheld videos with a forward view in visible light, which restricts the applicability of stabilization to domains such as UAV nighttime remote sensing. To fill this gap, we introduce a new multimodal UAV aerial video dataset (UAV-Test). Experiments show that our method consistently outperforms state-of-the-art online stabilizers in both quantitative metrics and visual quality, while achieving performance comparable to offline methods.