Structure-Preserving Deraining with Residue Channel Prior Guidance

Yi, Qiaosi; Li, Juncheng; Dai, Qinyan; Fang, Faming; Zhang, Guixu; Zeng, Tieyong

doi:10.1109/iccv48922.2021.00420

Cited by 60 publications

(23 citation statements)

References 44 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recently, many deep-learning based approaches [34,13,29,33,18,7,4,25,28,35] have been proposed for rain streak removal and made significant progress in this area. For example, Fu et al first adopt a shallow CNN [6] and then a deeper ResNet [7] to remove rain streaks.…”

Section: Cnn-based Methodsmentioning

confidence: 99%

“…Based on our proposed real and synthetic datasets RealRain-1k and SynRain-13k, we benchmark thirteen representative SID methods including two traditional ones [23,19] and eleven state-of-theart deep learning-based ones [37,36,3,35,4,13,29,18,8,6,7]. Specifically, we conduct the experiments on three tracks, i.e., (1) supervised learning (SL) track, (2) domain generalization (DG) track, and (3) cross-domain transfer learning (TL) track.…”

Section: Methodsmentioning

confidence: 99%

“…The aim of single image deraining (SID) is to restore the clean and rain-free background images from the rainy ones. Benefitting from the rapid development of deep convolutional neural networks (CNNs) [16], recent years have witnessed an explosive spread of using CNN models to perform SID [36,37,35,3,4,29,18,7]. they contribute to advancing the research for real-world SID, the difficulty in collecting high-quality rainy and clean pairs in the real world leads to various issues, including low resolution (256 × 256), homogeneous rain streaks, limited background variation (Fig.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Toward Real-world Single Image Deraining: A New Benchmark and Beyond

Li¹,

Zhang²,

Zhang³

et al. 2022

Preprint

View full text Add to dashboard Cite

Single image deraining (SID) in real scenarios attracts increasing attention in recent years. Due to the difficulty in obtaining real-world rainy/clean image pairs, previous real datasets suffer from low-resolution images, homogeneous rain streaks, limited background variation, and even misalignment of image pairs, resulting in incomprehensive evaluation of SID methods. To address these issues, we establish a new high-quality dataset named RealRain-1k, consisting of 1, 120 high-resolution paired clean and rainy images with low-and high-density rain streaks, respectively. Images in RealRain-1k are automatically generated from a large number of real-world rainy video clips through a simple yet effective rain density-controllable filtering method, and have good properties of high image resolution, background diversity, rain streaks variety, and strict spatial alignment. RealRain-1k also provides abundant rain streak layers as a byproduct, enabling us to build a large-scale synthetic dataset named SynRain-13k by pasting the rain streak layers on abundant natural images. Based on them and existing datasets, we benchmark more than 10 representative SID methods on three tracks: (1) fully supervised learning on RealRain-1k, (2) domain generalization to real datasets, and (3) syn-to-real transfer learning. The experimental results (1) show the difference of representative methods in image restoration performance and model complexity, (2) validate the significance of the proposed datasets for model generalization, and (3) provide useful insights on the superiority of learning from diverse domains and shed lights on the future research on real-world SID. The datasets will be released at https://github.com/hiker-lw/RealRain-1k.Recently, many research efforts have been made to solve SID from synthetic images to the real-world rainy images, where two real datasets are proposed, i.e., and RainDS [24]. While * Equal contribution.

show abstract

Section: Cnn-based Methodsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Toward Real-world Single Image Deraining: A New Benchmark and Beyond

Li¹,

Zhang²,

Zhang³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…This method makes the trained network model more robust to rain maps of natural scenes. In 2021 Yi et al [37] designed a deraining network with RCP guidance. The network can generate clean, high-quality, rain-free images directly regardless of any rainfall assumptions.…”

Section: Deep Network-based Deraining Methodsmentioning

confidence: 99%

Continuous learning deraining network based on residual FFT convolution and contextual transformer module

Zhang

Peng

et al. 2022

IET Image Processing

View full text Add to dashboard Cite

Single image deraining work has attracted a lot of attention. So an effective single image deraining network model, RFCTNet, is designed. Specifically, the residual Fast Fourier Transform convolution module is used as a baseline for the deraining network. The advantage is its ability to capture both long and short‐term information interactions and to integrate high and low‐frequency background information for delivery. And a novel Transformer module, the Contextual Transformer module, is introduced. Contextual knowledge of the rain streaks neighbourhood is learned to improve the recovery of textures and information in the image background based on focusing on the rain pattern features. Because of the overlap between rain streaks, removing them in one stage is difficult, so a multi‐stage progressive rain removal method is used. Considering the catastrophic forgetting problem of deep learning, a continuous learning method, PIGWM, is adopted, which is c and enables the network model to memorize of the previous data set. Also, the rain removal method RFCTNet has been proven through extensive experiments to outperform other advanced methods' evaluation metrics on both synthetic and real datasets. The source code can be found on https://github.com/SUTwu/RFCTNet

show abstract

“…recurrent architecture [19], [31], [32], [34], prior knowledgeguided model [2], [28], [48], generative network [1], [29], [40], [44], multi-scale pyramid [8], [16], [19], [38], [55], multistage learning [5], [21], [45], [47] and etc.. On the one hand, most synthetic datasets [7], [41], [49] for training the network do not consider veiling effect. And on the other hand, most existing methods [8], [16], [32], [38], [43] do not take detail reconstruction for rain removal into consideration.…”

Section: Related Workmentioning

confidence: 99%

From Heavy Rain Removal to Detail Restoration: A Faster and Better Network

Gao¹,

Wen²,

Zhang³

et al. 2022

Preprint

View full text Add to dashboard Cite

The dense rain accumulation in heavy rain can significantly wash out images and thus destroy the background details of images. Although existing deep rain removal models lead to improved performance for heavy rain removal, we find that most of them ignore the detail reconstruction accuracy of rain-free images. In this paper, we propose a dual-stage progressive enhancement network (DPENet) to achieve effective deraining with structure-accurate rain-free images. Two main modules are included in our framework, namely a rain streaks removal network (R 2 Net) and a detail reconstruction network (DRNet). The former aims to achieve accurate rain removal, and the latter is designed to recover the details of rain-free images. We introduce two main strategies within our networks to achieve trade-off between the effectiveness of deraining and the detail restoration of rain-free images. Firstly, a dilated dense residual block (DDRB) within the rain streaks removal network is presented to aggregate high/low level features of heavy rain. Secondly, an enhanced residual pixel-wise attention block (ERPAB) within the detail reconstruction network is designed for context information aggregation. We also propose a comprehensive loss function to highlight the marginal and regional accuracy of rainfree images. Extensive experiments on benchmark public datasets show both efficiency and effectiveness of the proposed method in achieving structure-preserving rain-free images for heavy rain removal. The source code and pre-trained models can be found at https://github.com/wybchd/DPENet.

show abstract

Structure-Preserving Deraining with Residue Channel Prior Guidance

Cited by 60 publications

References 44 publications

Toward Real-world Single Image Deraining: A New Benchmark and Beyond

Toward Real-world Single Image Deraining: A New Benchmark and Beyond

Continuous learning deraining network based on residual FFT convolution and contextual transformer module

From Heavy Rain Removal to Detail Restoration: A Faster and Better Network

Contact Info

Product

Resources

About