Edge-Informed Single Image Super-Resolution

Nazeri, Kamyar; Thasarathan, Harrish; Ebrahimi, Mehran

doi:10.1109/iccvw.2019.00409

Cited by 44 publications

(22 citation statements)

References 39 publications

(54 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In conventional methods, edge-based models [7]- [14] enhance the sharpness of super-resolved image by utilizing edge statistics. They model edge statistical dependencies by estimating the structural connectivity between HR and LR.…”

Section: A Related Workmentioning

confidence: 99%

“…However, they leave structural distortion problems and fail to recover details such as texture. To overcome the structural limitations of images , some models [13], [14] utilize structural information by designing additional modules for preserving the structure. Ma et al [13] utilized edge information in perceptual-driven methods by feeding explicit guidance to the established model, and Nazari et al [14] proposed an edge-informed SR method based on an image inpainting task.…”

Section: A Related Workmentioning

confidence: 99%

“…For high-fidelity image, it is necessary to represent such details as high frequency components such as texture and structural information. To address this issue, numerous SR methods have been proposed, such as conventional methods [7]- [14], deep learning methods [15]- [23] and the perceptual-driven method [3], [24]- [28].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Edge Profile Super Resolution

Lee

Yun

Kim³

2021

IEEE Access

View full text Add to dashboard Cite

The purpose of single-image super resolution (SISR) is to reconstruct an accurate highresolution image from a degraded low-resolution image. Owing to the lack of information in low-resolution images, SISR is a challenging problem. In particular, it is difficult to represent details, including highfrequency components, such as texture and structural information. We propose the edge profile superresolution (EPSR) method to preserve structural information and restore texture. EPSR is achieved by stacking modified fractal residual network (mFRN) structures hierarchically and repeatedly. Each mFRN is composed of many residual edge profile blocks (REPBs) that extract features and preserve the edge, structure, and texture information of the image. For implementing REPB, we design three main modules: Residual Efficient Channel Attention Block(RECAB) module, Edge Profile(EP) module, and Context Network(CN) module. By repeating the procedure in the mFRN structure, the EPSR method could be used to extract high-fidelity features, thus preventing texture loss and preserving the structure with appropriate sharpness. Experimental results show that EPSR achieves competitive performance against state-of-the-art methods in terms of the peak signal-to-noise ratio(PSNR) and structural similarity index measure(SSIM) evaluation metrics, as well as visual results.

show abstract

Section: A Related Workmentioning

confidence: 99%

Section: A Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Edge Profile Super Resolution

Lee

Yun

Kim³

2021

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Edge Guided Image Generation. Edge maps are usually adopted in image inpainting [47,42,35] and image super-resolution [43] tasks to reconstruct the missing structure information of the inputs. For example, Nazeri et al [42] propose an edge generator to hallucinate edges in the missing regions given edges, which can be regarded as an edge completion problem.…”

Section: Related Workmentioning

confidence: 99%

Edge Guided GANs with Contrastive Learning for Semantic Image Synthesis

Tang¹,

Qi²,

Guolei³

et al. 2020

Preprint

View full text Add to dashboard Cite

We propose a novel Edge guided Generative Adversarial Network (EdgeGAN) for photo-realistic image synthesis from semantic layouts. Although considerable improvement has been achieved, the quality of synthesized images is far from satisfactory due to two largely unresolved challenges. First, the semantic labels do not provide detailed structural information, making it difficult to synthesize local details and structures. Second, the widely adopted CNN operations such as convolution, down-sampling and normalization usually cause spatial resolution loss and thus are unable to fully preserve the original semantic information, leading to semantically inconsistent results (e.g., missing small objects). To tackle the first challenge, we propose to use edge as an intermediate representation which is further adopted to guide image generation via a proposed attention guided edge transfer module. Edge information is produced by a convolutional generator and introduces detailed structure information. Further, to preserve the semantic information, we design an effective module to selectively highlight class-dependent feature maps according to the original semantic layout. Extensive experiments on two challenging datasets show that the proposed EdgeGAN can generate significantly better results than state-ofthe-art methods. The source code and trained models are available at https://github.com/Ha0Tang/EdgeGAN.

show abstract

“…Previously, state-of-art methods [7], [8], [21] utilized edge information to mitigate endemic distortion problems in a super-resolution task. Such tendencies prove that the utilization of edges is an intuitive solution to improve super-resolved images.…”

Section: Introductionmentioning

confidence: 99%

Edge and Identity Preserving Network for Face Super-Resolution

Kim,

Li,

Yun

et al. 2020

Preprint

View full text Add to dashboard Cite

Face super-resolution has become an indispensable part in security problems such as video surveillance and identification system, but the distortion in facial components is a main obstacle to overcoming the problems. To alleviate it, most stateof-the-arts have utilized facial priors by using deep networks. These methods require extra labels, longer training time, and larger computation memory. Thus, we propose a novel Edge and Identity Preserving Network for Face Super-Resolution Network, named as EIPNet, which minimizes the distortion by utilizing a lightweight edge block and identity information. Specifically, the edge block extracts perceptual edge information and concatenates it to original feature maps in multiple scales. This structure progressively provides edge information in reconstruction procedure to aggregate local and global structural information. Moreover, we define an identity loss function to preserve identification of super-resolved images. The identity loss function compares feature distributions between super-resolved images and target images to solve unlabeled classification problem. In addition, we propose a Luminance-Chrominance Error (LCE) to expand usage of image representation domain. The LCE method not only reduces the dependency of color information by dividing brightness and color components but also facilitates our network to reflect differences between Super-Resolution (SR) and High-Resolution (HR) images in multiple domains (RGB and YUV). The proposed methods facilitate our super-resolution network to elaborately restore facial components and generate enhanced 8× scaled super-resolution images with a lightweight network structure. Furthermore, our network can generate a 128 × 128 superresolved image with 215 fps on a GTX 1080Ti GPU. Extensive experimental results demonstrate that our network qualitatively and quantitatively outperforms state-of-the-art methods on two challenging datasets: CelebA and VGGFace2.

show abstract

Edge-Informed Single Image Super-Resolution

Cited by 44 publications

References 39 publications

Edge Profile Super Resolution

Edge Profile Super Resolution

Edge Guided GANs with Contrastive Learning for Semantic Image Synthesis

Edge and Identity Preserving Network for Face Super-Resolution

Contact Info

Product

Resources

About