Guided Frequency Separation Network for Real-World Super-Resolution

Zhou, Yuanbo; Deng, Wei; Tong, Tong; Gao, Qinquan

doi:10.1109/cvprw50498.2020.00222

Cited by 50 publications

(41 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Adversarial loss is designed to dominate the training process, different from SRGAN [32] and ESRGAN [34] which mainly use pixel-wise loss for supervised training. Such learning strategy -unpaired degradation and paired SR -can also be found in some later works, including frequency separation for real-world SR (FSSR) [24] and frequency separation SRGAN (FS-SRGAN) [68]. These two methods both apply adversarial loss only to high-frequency contents in order to relieve the difficulty in adversarial training.…”

Section: Implicit Degradation Modelling 71 Learning Data Distribution...mentioning

confidence: 97%

See 1 more Smart Citation

Blind Image Super-Resolution: A Survey and Beyond

Liu¹,

Liu²,

Gu³

et al. 2021

Preprint

View full text Add to dashboard Cite

Blind image super-resolution (SR), aiming to super-resolve low-resolution images with unknown degradation, has attracted increasing attention due to its significance in promoting real-world applications. Many novel and effective solutions have been proposed recently, especially with the powerful deep learning techniques. Despite years of efforts, it still remains as a challenging research problem. This paper serves as a systematic review on recent progress in blind image SR, and proposes a taxonomy to categorize existing methods into three different classes according to their ways of degradation modelling and the data used for solving the SR model. This taxonomy helps summarize and distinguish among existing methods. We hope to provide insights into current research states, as well as to reveal novel research directions worth exploring. In addition, we make a summary on commonly used datasets and previous competitions related to blind image SR. Last but not least, a comparison among different methods is provided with detailed analysis on their merits and demerits using both synthetic and real testing images.

show abstract

Section: Implicit Degradation Modelling 71 Learning Data Distribution...mentioning

confidence: 97%

“…16: Overall architectures of methods with implicit modelling for data distribution learning. (a)CinCGAN [8]; (b)Degradation GAN[67] and FSSR[24], FS-SRGAN[68]; (c)DASR[69]; (d)pseudo-supervision[70]. "D" represents discriminator network with GAN framework.…”

mentioning

confidence: 99%

Blind Image Super-Resolution: A Survey and Beyond

Liu¹,

Liu²,

Gu³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…They introduced an intermediate LR domain to divide the SR process into the first unsupervised stage and the second supervised stage. Zhou et al [16]…”

Section: Unsupervised Image Super-resolutionmentioning

confidence: 99%

“…To tackle these difficulties, several methods extract facial prior knowledge, such as landmarks [11,12], parsing maps [12,13], and facial attributes [14], to recover HR face images while preserving facial structures. Also, to overcome the lack of paired data, an intermediate LR domain [15,16] is introduced for the transformation from LR domain to HR domain. However, these methods still cannot reconstruct photo-realistic high-resolution face images, particularly in an unsupervised manner.…”

mentioning

confidence: 99%

Unsupervised face super-resolution via gradient enhancement and semantic guidance

Tang

Zhou

et al. 2021

Vis Comput

View full text Add to dashboard Cite

Face super-resolution aims to recover high-resolution face images with accurate geometric structures. Most of the conventional super-resolution methods are trained on paired data that is difficult to obtain in the real-world setting. Besides, these methods do not fully utilize facial prior knowledge for face super-resolution. To tackle these problems, we propose an end-to-end unsupervised face super-resolution network to super-resolve low-resolution face images. We propose a gradient enhancement branch and a semantic guidance mechanism. Specifically, the gradient enhancement branch reconstructs high-resolution gradient maps, under the restriction of two proposed gradient losses. Then the super-resolution network integrates features in both image and gradient space to super-resolve face images with geometric structure preservation. Moreover, the proposed semantic guidance mechanism, including a semantic-adaptive sharpen module and a semantic-guided discriminator, can reconstruct sharp edges and improve local details in different facial regions adaptively, under the guidance of semantic parsing maps. Qualitative and quantitative experiments demonstrate that our proposed method can reconstruct high-resolution face images with sharp edges and photo-realistic details, outperforming the state-of-the-art methods.

show abstract

“…Moreover, images in the wild might contain corruptions and artifacts that should be removed in the super-resolution process. Therefore, blind SR approaches [4,34,15,55,38] tackle unknown degradation operations, whereas unsupervised real-world approaches [28,33,5,53,22,39,60] tackle severe degradation or compression artifacts in lowresolution images. Since our approach aims at performing super-resolution robust to noise and artifacts, we focus on the real-world literature.…”

Section: Single Image Super-resolutionmentioning

confidence: 99%

Generalized Real-World Super-Resolution through Adversarial Robustness

Castillo¹,

Escobar²,

Pérez³

et al. 2021

Preprint

View full text Add to dashboard Cite

Real-world Super-Resolution (SR) has been traditionally tackled by first learning a specific degradation model that resembles the noise and corruption artifacts in lowresolution imagery. Thus, current methods lack generalization and lose their accuracy when tested on unseen types of corruption. In contrast to the traditional proposal, we present Robust Super-Resolution (RSR), a method that leverages the generalization capability of adversarial attacks to tackle real-world SR. Our novel framework poses a paradigm shift in the development of real-world SR methods. Instead of learning a dataset-specific degradation, we employ adversarial attacks to create difficult examples that target the model's weaknesses. Afterward, we use these adversarial examples during training to improve our model's capacity to process noisy inputs. We perform extensive experimentation on synthetic and real-world images and empirically demonstrate that our RSR method generalizes well across datasets without re-training for specific noise priors. By using a single robust model, we outperform state-of-theart specialized methods on real-world benchmarks.

show abstract

Guided Frequency Separation Network for Real-World Super-Resolution

Cited by 50 publications

References 21 publications

Blind Image Super-Resolution: A Survey and Beyond

Blind Image Super-Resolution: A Survey and Beyond

Unsupervised face super-resolution via gradient enhancement and semantic guidance

Generalized Real-World Super-Resolution through Adversarial Robustness

Contact Info

Product

Resources

About