Recursive modified dense network for single-image deraining

Guoqiang, Chai; Wang, Zhaoba; Guo, Guodong; Chen, Youxing; Jin, Yingji; Wang, Wei; Zhao, Xia

doi:10.1117/1.jei.29.3.033006

Cited by 2 publications

(3 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Qian et al [26] synthesized a training dataset covered by raindrops and remove them from the background by an attention GAN with a multiscale objective function. Chai et al [27], [44] proposed two recurrent networks to remove rain streaks stage by stage. Yasarla and Patel [28] adopted an uncertainty guided network to detect rain streaks and then remove them with cycle spinning enhancement.…”

Section: B Single-image Based Methodsmentioning

confidence: 99%

“…Besides, a NLNM is applied after the autoencoder for a better representation and a convolution layer without any activation function is added at the end to generate the deraining image. As instance normalization (IN) and batch normalization (BN) may generate unpleasant artifacts [17,27], we apply spectral normalization (SN) for all convolution layers to enhance the training stability while improving the deraining performance.…”

Section: ) Symmetrical Autoencoder Modulementioning

confidence: 99%

See 1 more Smart Citation

Enhanced Attentive Generative Adversarial Network for Single-Image Deraining

et al. 2021

Self Cite

View full text Add to dashboard Cite

The problem of image rain removal has drawn widespread attention as the blurry images caused by rain streaks can degrade the performance of many computer vision tasks. Although exiting deep learning-based methods outperform most traditional methods, there are still unresolved issues in terms of performance. In this paper, we propose a novel enhanced attentive generative adversarial network named EAGAN to effectively remove the rain streaks and restore the image structural details at the same time. As rain streaks have different sizes and shapes, EAGAN utilizes a multiscale aggregation attention module (MAAM) to produce an attention map to guide the subsequent network to put conscious attention to rain regions. A symmetrical autoencoder with long-range skip-connections, squeeze-and-excitation (SE) modules, and non-local operation is further utilized to enhance the representation of the network. Finally, spectral normalization and a relativistic generative adversarial network (GAN) are further applied to improve the training stability and deraining performance. Both qualitative and quantitative validations on synthetic and real-world datasets demonstrate that the proposed approach can achieve a competitive performance in comparison with the state-of-the-art methods.INDEX TERMS Image deraining, attentive module, relativistic GAN.

show abstract

Section: B Single-image Based Methodsmentioning

confidence: 99%

Section: ) Symmetrical Autoencoder Modulementioning

confidence: 99%

Enhanced Attentive Generative Adversarial Network for Single-Image Deraining

et al. 2021

Self Cite

View full text Add to dashboard Cite

show abstract

“…The contextual information is very important to remove the rain pattern due to the availability of rain patterns with different sizes and shapes. This information is extracted by the recursive modified dense network through batch normalization layers (Chen and Wang, 2020). The network depth increases automatically, which affects the complexity of the deraining process.…”

mentioning

confidence: 99%

Performance analysis of a dual stage deep rain streak removal convolution neural network module with a modified deep residual dense network

Jayaraman,

Chinnusamy

2022

View full text Add to dashboard Cite

The visual appearance of outdoor captured images is affected by various weather conditions, such as rain patterns, haze, fog and snow. The rain pattern creates more degradation in the visual quality of the image due to its physical structure compared with other weather conditions. Also, the rain pattern affects both foreground and background image information. The removal of rain patterns from a single image is a critical process, and more attention is given to remove the structural rain pattern from real-time rain images. In this paper, we analyze the single image deraining problem and present a solution using the dual stage deep rain streak removal convolutional neural network. The proposed single image deraining framework primarily consists of three main blocks: a derain streaks removal CNN (derain SRCNN), a modified residual dense block (MRDB), and a six-stage scale feature aggregation module (3SFAM). The ablation study is conducted to evaluate the performance of various modules available in the proposed deraining network. The robustness of the proposed deraining network is evaluated over the popular synthetic and real-time data sets using four performance metrics such as the peak signal-to-noise ratio (PSNR), the feature similarity index (FSIM), the structural similarity index measure (SSIM), and the universal image quality index (UIQI). The experimental results show that the proposed framework outperforms both synthetic and real-time images compared with other state-of-the-art single image deraining approaches. In addition, the proposed network takes less running and training time.

show abstract

Recursive modified dense network for single-image deraining

Cited by 2 publications

References 0 publications

Enhanced Attentive Generative Adversarial Network for Single-Image Deraining

Enhanced Attentive Generative Adversarial Network for Single-Image Deraining

Performance analysis of a dual stage deep rain streak removal convolution neural network module with a modified deep residual dense network

Contact Info

Product

Resources

About