Overwater Image Dehazing via Cycle-Consistent Generative Adversarial Network

Zheng, Shunyuan; Sun, Jiamin; Liu, Qinglin; Qi, Yuankai; Yan, Jianen

doi:10.3390/electronics9111877

Cited by 8 publications

(3 citation statements)

References 57 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In order to improve the image quality and diversity generated by GANs, the current research on GANs almost stays in the spatial domain. For example, CycleGAN++ [30] improves the phenomenon that the CycleGAN [13] generator may hide some features in the input image and restore them at the time of output, eliminating the circular structure of CycleGAN, removing the loss of cyclic consistency and adding the loss of classification; StyleGAN2 [31] improved the problem that the image generated by StyleGAN [14] may have obvious water droplets. These are improvements to some problems existing in the airspace of the images generated by the existing GANs Although there are also improvements for frequency-domain characteristics, the optimization methods are usually different for different frequency-domain characteristics.…”

Section: High-order Optimization Of Image Generationmentioning

confidence: 99%

Frequency domain characteristics and optimization of image generation for GANs

Ye,

Sui,

Zeng

et al. 2024

Preprint

View full text Add to dashboard Cite

In recent years, Generative Adversarial Networks (GANs) have demonstrated enormous promise in areas connected to image generation. As the model generation performance continues to improve and the generated images become more realistic, it is difficult to effectively distinguish between the real image and the generated image. Therefore, the problem of discriminating and optimizing the generated images (adversarial discrimination) has become necessary, and subsequent optimization plans are proposed based on the discrimination strategy. However, due to the nature of convolution, the two-dimensional power spectrum curve of the generated image is low overall; that is, compared with the real image, there is energy loss at each frequency (without other processing), and the curve drops rapidly and approaches zero, which is obviously different from the real image. In particular, the curve of the image generated by transposed convolution has a clear upward trend at the very high-frequency part, which is contrary to the characteristic of the real image, which is that the energy decreases with increasing frequency. Based on the discussion of the characteristics and inducements of the two-dimensional power spectrum curve of the generated image, we present a discrimination approach based on curve warping at high frequency and energy loss to improve the discrimination capacity of the generated image and realize the effective discrimination between the real image and the generated image. Based on this, we present the power spectrum loss function to improve the upward warping characteristics of the very high-frequency part of the two-dimensional power spectrum curve without degrading the quality of the generated image and the high-frequency feature loss function to improve the quality of the generated image. The value and efficiency of the proposed discrimination approach in this study are demonstrated on multiple GANs models, including WGAN, WGAN-GP, and SAGAN, with the dataset celeba, and the GANs model with encoder-decoder as the generator with the dataset CelebA-HQ. The two loss functions proposed are also demonstrated on multiple GANs models, including WGAN, WGAN-GP, and SAGAN with the dataset FFHQ. After adding the high-frequency feature loss, the FID decreases by 5.97, 5.15, and 6.56, respectively. After adding the power spectrum loss, the above models can improve the upward warping characteristics of the two-dimensional power spectrum curve in the very high-frequency part of the generated image to a certain extent. The FID decreases by 17.4, 11.55 and 12.27 when the weight is fixed, and 12.66, 8.15 and 4.46 when the weight is variable, respectively.

show abstract

Section: High-order Optimization Of Image Generationmentioning

confidence: 99%

Frequency domain characteristics and optimization of image generation for GANs

Ye,

Sui,

Zeng

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…In 2023, Zhang et al, [23] suggested an enhanced cycle-consistent attacker network-based imagedefogging system. Then, to enhance the network's capacity for feature extraction, the self-recognition module and the multi-scale feature fusion module for atrous convolution are built on the conventional Cycle GAN network.…”

Section: Literature Surveymentioning

confidence: 99%

Deep-APT: Deep Learning based Efficient Accident Prevention Technique in Fogged Environment

Lokesh

Ahilan³

2023

Preprint

View full text Add to dashboard Cite

Image defogging is an innovative strategy for recovering images in foggy environments that has gotten a lot of attention in recent years because of its use in surveillance systems. The standard defogging algorithm, on the other hand, has difficulty merging the depth of picture detail and the colour of the picture. In this paper, a novel Accident Prevention Technique (Deep-APT) has been proposed to effectively restore fog-free images and prevent accidents using FasterRCNN network. Initially, a dashboard camera monitors the road ahead of the vehicle and collects video. This video sequence is converted to frames. The transformed images are pre-processed using an Adaptive dual threshold Tetrolet transform that preprocess foggy images to fog-free images it is used to remove noise in the input image. Based on the defogged image, use FasterRCNN technology to detect objects in front of the car. The Deep-APT method has been simulated using MATLAB. The experimental result shows the proposed Deep-APT yields an overall accuracy is 99.52%. As compared to existing techniques, the proposed FasterRCNN network shows better results in terms of precision, F1 score, accuracy, and recall. Using DAWN dataset, the MSE, SSIM and PSNR values for the proposed method are 0.12, 0.65 and 0.12. The Deep-APT network improves the overall accuracy of 15.43%, and 4.72% better than CR-YOLnet, and RDL respectively.

show abstract

“…This dynamic nature poses challenges for image restoration, as algorithms must adapt to evolving weather conditions. Thus, effective image restoration techniques tailored to the unique characteristics of maritime environments are essential for ensuring safe and efficient maritime activities (Zheng et al, 2020). This raises a question: how to better help image restoration in adapting to the complex and ever-changing maritime scenes under adverse weather conditions?…”

mentioning

confidence: 99%

Learning degradation-aware visual prompt for maritime image restoration under adverse weather conditions

He,

Jia,

2024

Front. Mar. Sci.

View full text Add to dashboard Cite

Adverse weather conditions such as rain and haze often lead to a degradation in the quality of maritime images, which is crucial for activities like navigation, fishing, and search and rescue. Therefore, it is of great interest to develop an effective algorithm to recover high-quality maritime images under adverse weather conditions. This paper proposes a prompt-based learning method with degradation perception for maritime image restoration, which contains two key components: a restoration module and a prompting module. The former is employed for image restoration, whereas the latter encodes weather-related degradation-specific information to modulate the restoration module, enhancing the recovery process for improved results. Inspired by the recent trend of prompt learning in artificial intelligence, this paper adopts soft-prompt technology to generate learnable visual prompt parameters for better perceiving the degradation-conditioned cues. Extensive experimental results on several benchmarks show that our approach achieves superior restoration performance in maritime image dehazing and deraining tasks.

show abstract

Overwater Image Dehazing via Cycle-Consistent Generative Adversarial Network

Cited by 8 publications

References 57 publications

Frequency domain characteristics and optimization of image generation for GANs

Frequency domain characteristics and optimization of image generation for GANs

Deep-APT: Deep Learning based Efficient Accident Prevention Technique in Fogged Environment

Learning degradation-aware visual prompt for maritime image restoration under adverse weather conditions

Contact Info

Product

Resources

About