Towards Real-World Adverse Weather Image Restoration: Enhancing Clearness and Semantics with Vision-Language Models

TL;DR

This paper proposes a semi-supervised framework using vision-language models to improve real-world adverse weather image restoration, addressing synthetic data limitations and enhancing clearness and semantics.

Contribution

It introduces a novel semi-supervised learning approach leveraging vision-language models for real-world weather image restoration, with a dual-step strategy for clearness and semantic enhancement.

Findings

Achieves superior restoration quality on real-world adverse weather images.

Outperforms state-of-the-art methods in qualitative and quantitative evaluations.

Effectively utilizes real data for training through pseudo-labels and weather prompt learning.

Abstract

This paper addresses the limitations of adverse weather image restoration approaches trained on synthetic data when applied to real-world scenarios. We formulate a semi-supervised learning framework employing vision-language models to enhance restoration performance across diverse adverse weather conditions in real-world settings. Our approach involves assessing image clearness and providing semantics using vision-language models on real data, serving as supervision signals for training restoration models. For clearness enhancement, we use real-world data, utilizing a dual-step strategy with pseudo-labels assessed by vision-language models and weather prompt learning. For semantic enhancement, we integrate real-world data by adjusting weather conditions in vision-language model descriptions while preserving semantic meaning. Additionally, we introduce an effective training strategy to bootstrap restoration performance. Our approach achieves superior results in real-world adverse weather image restoration, demonstrated through qualitative and quantitative comparisons with state-of-the-art works.