Contrastive Learning for Compact Single Image Dehazing

By Haiyan Wu, Yanyun Qu, Shaohui Lin, Jian Zhou, Ruizhi Qiao, Zhizhong Zhang, Yuan Xie, Lizhuang Ma. Computer Vision and Pattern Recognition, 2021

860 citations recorded by the source corpus.

Single image dehazing is a challenging ill-posed problem due to the severe information degeneration. However, existing deep learning based dehazing methods only adopt clear images as positive samples to guide the training of dehazing network while negative information is unexploited. Moreover, most of them focus on strengthening the dehazing network with an increase of depth and width, leading to a significant requirement of computation and memory. In this paper, we propose a novel contrastive regularization (CR) built upon contrastive learning to exploit both the information of hazy images and clear images as negative and positive samples, respectively. CR ensures that the restored image is pulled to closer to the clear image and pushed to far away from the hazy image in the representation space.Furthermore, considering trade-off between performance and memory storage, we develop a compact dehazing network based on autoencoder-like (AE) framework. It involves an adaptive mixup operation and a dynamic feature enhancement module, which can benefit from preserving information flow adaptively and expanding the receptive field to improve the network’s transformation capability, respectively. We term our dehazing network with autoencoder and contrastive regularization as AECR-Net. The extensive experiments on synthetic and real-world datasets demonstrate that our AECR-Net surpass the state-of-the-art approaches. The code is released in https://github.com/GlassyWu/AECR-Net.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/fc5ed765f8a07d265936994a3ef2b2ac441e7283

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Dacheng Tao
2. Chongyi Li
3. Vishal M. Patel
4. Chunle Guo
5. Jing Zhang
6. Wenqi Ren
7. Runmin Cong
8. Saeed Anwar

Citations in the corpus

• Image Dehazing Transformer with Transmission-Aware 3D Position Embedding (2022, methodology): “Following previous works [10,20,30], we use ITS and OTS subsets of RESIDE dataset [18] as the training datasets and conduct the evaluations on SOTS subset that contains 500 indoor and 500 outdoor hazy images.”
• DEA-Net: Single Image Dehazing Based on Detail-Enhanced Convolution and Content-Guided Attention (2023, methodology): “We observe that fusing the features from the encoder part with that from the decoder part is an effective trick in dehazing and other low-level vision tasks [6], [10], [38], [39].”
• Stripformer: Strip Transformer for Fast Image Deblurring (2022, methodology): “Recently, it has been adopted in vision tasks [14,31] by pulling close “positive” (similar) pairs and pushing apart “negative” (dissimilar) pairs in the feature space.”
• A Survey on Self-Supervised Learning: Algorithms, Applications, and Future Trends (2023, methodology): “Numerous CL-based methods are available in the literature [89]–[109].”
• Dual Temperature Helps Contrastive Learning Without Many Negative Samples: Towards Understanding and Simplifying MoCo (2022, background): “CL remains as a mainstream framework for SSL and has also been extensively studied in other filed applications [1, 13, 16, 27, 37, 46, 51, 54, 55, 58, 64].”
• TransWeather: Transformer-based Restoration of Images Degraded by Adverse Weather Conditions (2021, background): “Recently, Convolutional Neural Networks (CNNs) based solutions have been explored extensively for deraining [11, 37, 53, 56, 61, 63, 69, 70, 76], dehazing [8, 19, 41, 57, 69, 71, 72], desnowing [29, 44, 73] and raindrop removal [37, 40, 66].”
• Vision Transformers for Single Image Dehazing (2022, methodology): “Comparing the baseline methods, FFA-Net and AECR-Net are far superior to the previous or contemporaneous methods.”
• All-In-One Image Restoration for Unknown Corruption (2022, background): “In recent, some studies have shown the effectiveness of contrastive learning in image restorations [43, 47].”