Natural and Realistic Single Image Super-Resolution With Explicit Natural Manifold Discrimination

Soh, Jae Woong; Park, Gu Yong; Jo, Jung Hyun; Cho, Nam Ik

doi:10.1109/cvpr.2019.00831

Cited by 110 publications

(71 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As expected, the PSNR values obtained by PSNR-driven methods are higher than these of GAN-based methods, but the PI indicators correlating with perceptual quality universally far lag behind these of GAN-based methods. The SRLRGAN-SN method wins the first place in SOAT methods regarding to average PI, where the mean value is 3.14, 2.79, 0.24, 0.36, 0.37, 0.21 and 0.58 lower than that of VDSR [16], EDSR [17], SRGAN [22], EnhanceNet [23], ESRGAN [24], PESRGAN [25] and NatSRGAN [26] separately. Compared with PSNR-driven methods, Our proposed method also achieves comparable performance in terms of PSNR.…”

Section: Comparison Of Other Popular Methodsmentioning

confidence: 95%

“…All results are achieved on Set5, Set14, BSD100 and Urban100 dataset [22], respectively. The SOAT methods include VDSR [16], EDSR [17], SRGAN [22], EnhanceNet [23], ESRGAN [24], PESRGAN [25], NatSRGAN [26] For perceptual super-resolution imaging, good perceptual quality is crucial from a perspective of human vision. In the aspect of quantitative metric, we use PI in equations (1) couple with PSNR to evaluate the objective performance [45], although the PSNR is not as effective as the PI metric in terms of perceptual quality.…”

Section: Comparison Of Other Popular Methodsmentioning

confidence: 99%

“…Vu et al [25] employ the relativistic discriminator and put forward the PESRGAN model coupling with focal loss and total variance loss. Soh et al [26] put forward the NatSRGAN model through the use of the natural manifold discriminator. In addition to the typical reconstruction loss, naturalness loss and relativistic adversarial loss are added in the NatSRGAN to further encourage images towards more natural manifold.…”

Section: A Super-resolution For Imagingmentioning

confidence: 99%

“…To prove the computational ability of SRLRGAN-SN, a study about the number of FLOP and the parameters is conducted. It is compared with the VDSR [16], EDSR [17], SRGAN [22], EnhanceNet [23], ESRGAN [24], PESRGAN [25], NatSRGAN [26]. Results of complexity analysis are depicted in Table VIII.…”

Section: E Complexity Analysismentioning

confidence: 99%

See 3 more Smart Citations

Least Squares Relativistic Generative Adversarial Network for Perceptual Super-Resolution Imaging

2020

View full text Add to dashboard Cite

Currently, deep-learning-based methods have been the most popular super-resolution techniques owing to the improvement of super-resolution performance. However, they are still lack perceptual fine details and thus result in unsatisfying visual quality. This paper proposes a novel method for high-quality perceptual super-resolution imaging, named SRLRGAN-SN. It aims to recovery visually plausible images with perceptual texture details by using the least squares relativistic generative adversarial network (GAN). The method applies the spectral normalization on the network with the target of enhancing the performance of GAN for super-resolution task. The least squares relativistic discriminator is designed to drive reconstruction images approximating high-quality perceptual manifold. Besides, a novel perceptual loss assembly is proposed to preserve structural texture details as much as possible. Results of experiment show that our method can not only recovery more visually realistic details, but also outperforms other popular methods regarding to quantitative metrics and perceptual evaluations. INDEX TERMS Generative adversarial network, super-resolution imaging, relativistic discriminator, perceptual quality, spectral normalization.

show abstract

Section: Comparison Of Other Popular Methodsmentioning

confidence: 95%

Section: Comparison Of Other Popular Methodsmentioning

confidence: 99%

Section: A Super-resolution For Imagingmentioning

confidence: 99%

Section: E Complexity Analysismentioning

confidence: 99%

See 2 more Smart Citations

Least Squares Relativistic Generative Adversarial Network for Perceptual Super-Resolution Imaging

2020

View full text Add to dashboard Cite

show abstract

“…The relativistic discriminator proposed by the relativistic GAN [27,28] shows that the effect of the relativistic discriminator can be improved by wrapping a focal loss [34]. NatSR [37] remodeled super-resolution, improved the network's discriminator, and improved the perceptual quality of SR images. The proposed second-order statistical feature can improve the discriminative representation of the network.…”

Section: Loss Functionmentioning

confidence: 99%

Generative Adversarial Network-Based Super-Resolution Considering Quantitative and Perceptual Quality

Wang

Cheng

et al. 2020

Symmetry

View full text Add to dashboard Cite

In recent years, the common algorithms for image super-resolution based on deep learning have been increasingly successful, but there is still a large gap between the results generated by each algorithm and the ground-truth. Even some algorithms that are dedicated to image perception produce more textures that do not exist in the original image, and these artefacts also affect the visual perceptual quality of the image. We believe that in the existing perceptual-based image super-resolution algorithm, it is necessary to consider Super-Resolution (SR) image quality, which can restore the important structural parts of the original picture. This paper mainly improves the Enhanced Super-Resolution Generative Adversarial Networks (ESRGAN) algorithm in the following aspects: adding a shallow network structure, adding the dual attention mechanism in the generator and the discriminator, including the second-order channel mechanism and spatial attention mechanism and optimizing perceptual loss by adding second-order covariance normalization at the end of feature extractor. The results of this paper ensure image perceptual quality while reducing image distortion and artefacts, improving the perceived similarity of images and making the images more in line with human visual perception.

show abstract

Multiscale generative adversarial network for real‐world super‐resolution

Sun

Yang

Tao

et al. 2021

Concurrency and Computation

View full text Add to dashboard Cite

Summary Recently, most deep convolutional neural networks used for image super‐resolution have achieved impressive performance on ideal datasets. However, these methods always fail in real‐world super‐resolution, and the results are blurred and structurally deformed. In this paper, a multiscale generative adversarial network (MGAN) is proposed to alleviate these issues. The model's multiscale loss function can effectively reduce the solution space and obtain the best features to reconstruct the image. The degraded framework based on kernel estimation and noise injection is mainly applied to obtain LR images that share the same domain with real‐world pictures. Moreover, the gradient branch is presented to provide other structural priors for SR processing. Simultaneously, to obtain better visual effects, LPIPS is used for perceptual losses instead of Visual Geometry Group (VGG). The competitive results show that our MGAN model outperforms the state‐of‐the‐art methods, resulting in lower noise and better visual quality, and reflects the superiority in image structure restoration.

show abstract

Natural and Realistic Single Image Super-Resolution With Explicit Natural Manifold Discrimination

Cited by 110 publications

References 24 publications

Least Squares Relativistic Generative Adversarial Network for Perceptual Super-Resolution Imaging

Least Squares Relativistic Generative Adversarial Network for Perceptual Super-Resolution Imaging

Generative Adversarial Network-Based Super-Resolution Considering Quantitative and Perceptual Quality

Multiscale generative adversarial network for real‐world super‐resolution

Contact Info

Product

Resources

About