DRACO-DehazeNet: An Efficient Image Dehazing Network Combining Detail Recovery and a Novel Contrastive Learning Paradigm

2024-10-18

arXiv.org

AI Summary

Existing learning-based image dehazing methods suffer from heavy reliance on large-scale labeled datasets, high computational overhead, and poor generalization under non-uniform or dense haze conditions. To address these limitations, this paper proposes a few-shot efficient dehazing framework. Methodologically, we introduce a novel quaternary contrastive loss that jointly enforces discriminative feature alignment between hazy and haze-free images and penalizes quality discrepancies among sub-module outputs. We further design a densely dilated inverted residual block to enhance multi-scale contextual modeling and incorporate a context-aware attention mechanism to improve fine-detail recovery. Extensive experiments demonstrate state-of-the-art performance across multiple benchmark datasets. Our approach reduces training data requirements by approximately 60% and lowers FLOPs by 35%, while maintaining lightweight model architecture and strong cross-scenario generalization. The source code is publicly available.

Technology Category

• Computer Vision: Learning & Optimization for CV
• Search and Optimization: Learning to Search
• Machine Learning: Deep Neural Architectures and Foundation Models

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

Image dehazing is crucial for clarifying images obscured by haze or fog, but current learning-based approaches is dependent on large volumes of training data and hence consumed significant computational power. Additionally, their performance is often inadequate under non-uniform or heavy haze. To address these challenges, we developed the Detail Recovery And Contrastive DehazeNet, which facilitates efficient and effective dehazing via a dense dilated inverted residual block and an attention-based detail recovery network that tailors enhancements to specific dehazed scene contexts. A major innovation is its ability to train effectively with limited data, achieved through a novel quadruplet loss-based contrastive dehazing paradigm. This approach distinctly separates hazy and clear image features while also distinguish lower-quality and higher-quality dehazed images obtained from each sub-modules of our network, thereby refining the dehazing process to a larger extent. Extensive tests on a variety of benchmarked haze datasets demonstrated the superiority of our approach. The code repository for this work is available at https://github.com/GreedYLearner1146/DRACO-DehazeNet.

Problem

• Efficient image dehazing with limited training data
• Improved performance under non-uniform or heavy haze
• Novel contrastive learning for separating hazy and clear features

Innovation

• Dense dilated inverted residual block
• Attention-based detail recovery network
• Quadruplet loss-based contrastive learning