Scaling Up Single Image Dehazing Algorithm by Cross-Data Vision Alignment for Richer Representation Learning and Beyond

2024-07-20

AI Summary

Existing single-image dehazing methods suffer from weak generalization and insufficient feature robustness when jointly trained on multi-source data, primarily due to neglecting inter-domain discrepancies. To address this, we propose a cross-dataset visual alignment framework featuring a novel “dual alignment” mechanism: external alignment integrates heterogeneous datasets to enhance domain invariance, while internal self-supervised learning strengthens local detail modeling. This work establishes the first large-scale aligned training paradigm specifically designed for dehazing. Our approach synergistically combines self-supervised learning, cross-domain feature alignment, high-resolution representation modeling, and multi-source collaborative training. Evaluated on the Natural Image Dataset, it significantly narrows the domain gap and achieves state-of-the-art PSNR and SSIM scores—producing restored images with visual quality closest to ground-truth haze-free references.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

In recent years, deep neural networks tasks have increasingly relied on high-quality image inputs. With the development of high-resolution representation learning, the task of image dehazing has received significant attention. Previously, many methods collect diverse image data for large-scale training to boost the performance on a target scene. Ignoring the domain gap between different data, former de-hazing methods simply adopt multiple datasets for explicit large-scale training, which often makes the methods themselves be violated. To address this problem, we propose a novel method of cross-data vision alignment for richer representation learning to improve the existing dehazing methodology. Specifically, we call for the internal- and external knowledge should be further adapted with a self-supervised manner to fill up the domain gap. By using cross-data external alignment, the datasets inherit samples from different domains that are firmly aligned, making the model learn more robust and generalizable features. By using the internal augmentation method, the model can fully exploit local information within the images, and then obtaining more image details. To demonstrate the effectiveness of our proposed method, we conduct training on the Natural Image Dataset (NID). Experimental results show that our method clearly resolves the domain gap in different dehazing datasets and presents a new pipeline for large-scale training in the dehazing task. Our approach significantly outperforms other advanced methods in dehazing and produces dehazed images that are closest to real haze-free images.

Problem

• Addressing domain gaps in diverse dehazing datasets
• Enhancing representation learning via cross-data alignment
• Improving dehazing performance with self-supervised adaptation

Innovation

• Cross-data vision alignment for domain gap reduction
• Self-supervised internal and external knowledge adaptation
• Internal augmentation for detailed image exploitation