Blind Image Super Resolution Using Deep Unsupervised Learning

The goal of single image super resolution (SISR) is to recover a high-resolution (HR) image from a low-resolution (LR) image. Deep learning based methods have recently made a remarkable performance gain in terms of both the effectiveness and efficiency for SISR. Most existing methods have to be trai...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Kazuhiro Yamawaki, Yongqing Sun, Xian-Hua Han
Formato:	article
Lenguaje:	EN
Publicado:	MDPI AG 2021
Materias:	image super resolution blind unsupervised learning blur kernel learning generated network degradation operation Electronics TK7800-8360
Acceso en línea:	https://doaj.org/article/e69d84375f974e61a94f0ec7054d7e12
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:e69d84375f974e61a94f0ec7054d7e12
record_format	dspace
spelling	oai:doaj.org-article:e69d84375f974e61a94f0ec7054d7e122021-11-11T15:37:19ZBlind Image Super Resolution Using Deep Unsupervised Learning10.3390/electronics102125912079-9292https://doaj.org/article/e69d84375f974e61a94f0ec7054d7e122021-10-01T00:00:00Zhttps://www.mdpi.com/2079-9292/10/21/2591https://doaj.org/toc/2079-9292The goal of single image super resolution (SISR) is to recover a high-resolution (HR) image from a low-resolution (LR) image. Deep learning based methods have recently made a remarkable performance gain in terms of both the effectiveness and efficiency for SISR. Most existing methods have to be trained based on large-scale synthetic paired data in a fully supervised manner. With the available HR natural images, the corresponding LR images are usually synthesized with a simple fixed degradation operation, such as bicubic down-sampling. Then, the learned deep models with these training data usually face difficulty to be generalized to real scenarios with unknown and complicated degradation operations. This study exploits a novel blind image super-resolution framework using a deep unsupervised learning network. The proposed method can simultaneously predict the underlying HR image and its specific degradation operation from the observed LR image only without any prior knowledge. The experimental results on three benchmark datasets validate that our proposed method achieves a promising performance under the unknown degradation models.Kazuhiro YamawakiYongqing SunXian-Hua HanMDPI AGarticleimage super resolutionblind unsupervised learningblur kernel learninggenerated networkdegradation operationElectronicsTK7800-8360ENElectronics, Vol 10, Iss 2591, p 2591 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	image super resolution blind unsupervised learning blur kernel learning generated network degradation operation Electronics TK7800-8360
spellingShingle	image super resolution blind unsupervised learning blur kernel learning generated network degradation operation Electronics TK7800-8360 Kazuhiro Yamawaki Yongqing Sun Xian-Hua Han Blind Image Super Resolution Using Deep Unsupervised Learning
description	The goal of single image super resolution (SISR) is to recover a high-resolution (HR) image from a low-resolution (LR) image. Deep learning based methods have recently made a remarkable performance gain in terms of both the effectiveness and efficiency for SISR. Most existing methods have to be trained based on large-scale synthetic paired data in a fully supervised manner. With the available HR natural images, the corresponding LR images are usually synthesized with a simple fixed degradation operation, such as bicubic down-sampling. Then, the learned deep models with these training data usually face difficulty to be generalized to real scenarios with unknown and complicated degradation operations. This study exploits a novel blind image super-resolution framework using a deep unsupervised learning network. The proposed method can simultaneously predict the underlying HR image and its specific degradation operation from the observed LR image only without any prior knowledge. The experimental results on three benchmark datasets validate that our proposed method achieves a promising performance under the unknown degradation models.
format	article
author	Kazuhiro Yamawaki Yongqing Sun Xian-Hua Han
author_facet	Kazuhiro Yamawaki Yongqing Sun Xian-Hua Han
author_sort	Kazuhiro Yamawaki
title	Blind Image Super Resolution Using Deep Unsupervised Learning
title_short	Blind Image Super Resolution Using Deep Unsupervised Learning
title_full	Blind Image Super Resolution Using Deep Unsupervised Learning
title_fullStr	Blind Image Super Resolution Using Deep Unsupervised Learning
title_full_unstemmed	Blind Image Super Resolution Using Deep Unsupervised Learning
title_sort	blind image super resolution using deep unsupervised learning
publisher	MDPI AG
publishDate	2021
url	https://doaj.org/article/e69d84375f974e61a94f0ec7054d7e12
work_keys_str_mv	AT kazuhiroyamawaki blindimagesuperresolutionusingdeepunsupervisedlearning AT yongqingsun blindimagesuperresolutionusingdeepunsupervisedlearning AT xianhuahan blindimagesuperresolutionusingdeepunsupervisedlearning
_version_	1718434956509708288

Blind Image Super Resolution Using Deep Unsupervised Learning

Ejemplares similares