Continual Learning-Based Unified Model for Unpaired Image Restoration Tasks

2025-07-25

AI Summary

To address multi-type image degradations—such as those caused by fog, snow, and rain—this paper proposes the first unified image restoration framework tailored for continual learning. Methodologically, it introduces a selective kernel fusion layer to enable degradation-adaptive dynamic feature integration; incorporates Elastic Weight Consolidation (EWC) to mitigate catastrophic forgetting; and proposes a Cycle-Contrastive Loss to enhance semantic consistency and discriminability across weather domains without requiring paired data. Evaluated on multiple benchmark datasets, the end-to-end model significantly outperforms existing single-task and multi-task approaches in PSNR, SSIM, and perceptual quality—achieving state-of-the-art performance. These results validate the framework’s strong generalization capability and robust continual learning efficacy under evolving weather-related degradation scenarios.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Life-Long and Continual Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Restoration of images contaminated by different adverse weather conditions such as fog, snow, and rain is a challenging task due to the varying nature of the weather conditions. Most of the existing methods focus on any one particular weather conditions. However, for applications such as autonomous driving, a unified model is necessary to perform restoration of corrupted images due to different weather conditions. We propose a continual learning approach to propose a unified framework for image restoration. The proposed framework integrates three key innovations: (1) Selective Kernel Fusion layers that dynamically combine global and local features for robust adaptive feature selection; (2) Elastic Weight Consolidation (EWC) to enable continual learning and mitigate catastrophic forgetting across multiple restoration tasks; and (3) a novel Cycle-Contrastive Loss that enhances feature discrimination while preserving semantic consistency during domain translation. Further, we propose an unpaired image restoration approach to reduce the dependance of the proposed approach on the training data. Extensive experiments on standard benchmark datasets for dehazing, desnowing and deraining tasks demonstrate significant improvements in PSNR, SSIM, and perceptual quality over the state-of-the-art.

Problem

• Unified model for restoring images degraded by multiple weather conditions
• Addresses catastrophic forgetting in continual learning for image restoration
• Reduces dependency on paired training data for image restoration

Innovation

• Selective Kernel Fusion for adaptive features
• Elastic Weight Consolidation prevents forgetting
• Cycle-Contrastive Loss enhances feature discrimination