Cat-AIR: Content and Task-Aware All-in-One Image Restoration

2025-03-23

AI Summary

This paper addresses the challenging problem of generic single-image restoration under unknown mixed degradations—including blur, noise, and compression artifacts. Methodologically, it proposes a content- and task-aware adaptive framework: (i) a novel alternating spatial-channel attention mechanism that dynamically allocates computational resources based on both image content complexity and restoration task requirements; (ii) cross-layer channel attention and cross-feature-space attention to enhance multi-scale feature interaction; and (iii) a smooth continual learning strategy enabling incremental adaptation to new degradation types while mitigating catastrophic forgetting of previously learned tasks. Extensive experiments demonstrate state-of-the-art performance across diverse restoration tasks—including denoising, deblurring, and JPEG artifact reduction—while achieving significantly lower FLOPs than existing generic models. To our knowledge, this is the first work to unify high accuracy and high efficiency in a single, fully functional image restoration framework, establishing a new benchmark for efficient universal restoration.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Learning on the Edge & Model Compression
• Natural Language Processing: Generation

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

All-in-one image restoration seeks to recover high-quality images from various types of degradation using a single model, without prior knowledge of the corruption source. However, existing methods often struggle to effectively and efficiently handle multiple degradation types. We present Cat-AIR, a novel extbf{C}ontent extbf{A}nd extbf{T}ask-aware framework for extbf{A}ll-in-one extbf{I}mage extbf{R}estoration. Cat-AIR incorporates an alternating spatial-channel attention mechanism that adaptively balances the local and global information for different tasks. Specifically, we introduce cross-layer channel attentions and cross-feature spatial attentions that allocate computations based on content and task complexity. Furthermore, we propose a smooth learning strategy that allows for seamless adaptation to new restoration tasks while maintaining performance on existing ones. Extensive experiments demonstrate that Cat-AIR achieves state-of-the-art results across a wide range of restoration tasks, requiring fewer FLOPs than previous methods, establishing new benchmarks for efficient all-in-one image restoration.

Problem

• Handles multiple image degradations with one model
• Balances local and global information adaptively
• Achieves efficient restoration with fewer computations

Innovation

• Alternating spatial-channel attention mechanism
• Cross-layer and cross-feature attentions
• Smooth learning strategy for new tasks