Patch loss: A generic multi-scale perceptual loss for single image super-resolution

doi:10.1016/j.patcog.2023.109510

CORC > 自动化研究所 > 中国科学院自动化研究所 > 多模态人工智能系统全国重点实验室

	Patch loss: A generic multi-scale perceptual loss for single image super-resolution
	An, Tai 1,2; Mao, Binjie 1,2; Xue, Bin 1,2; Huo, Chunlei 1,2; Xiang, Shiming 1,2; Pan, Chunhong 1
刊名	PATTERN RECOGNITION
	2023-07-01
卷号	139 页码:12
关键词	Single -image super -resolution Multi -scale loss functions Image visual perception Perceptual metrics
ISSN号	0031-3203
DOI	10.1016/j.patcog.2023.109510
通讯作者	Huo, Chunlei(clhuo@nlpr.ia.ac.cn)
英文摘要	In single image super-resolution (SISR), although PSNR is a key metric for signal fidelity, images with high PSNR do not necessarily render high visual quality. As a result, current perception-driven SISR meth-ods employ perceptual metrics close to the human eye to measure the quality of the generated images. Unfortunately, the perceptual loss and adversarial loss, widely used by the perception-driven SISR meth-ods, still underperform on these non-differentiable perceptual metrics. To this end, we propose a generic multi-scale perceptual loss, i.e., the patch loss, which can be easily plugged into off-the-shelf SISR meth-ods to improve a broad range of perceptual metrics. Specifically, the proposed patch loss minimizes the multi-scale similarity of image patches and enhances the restoration of regions with complex textures and sharp edges via parameter-free adaptive patch-wise attention. Our proposed patch loss introduces more realistic details compared to the perceptual loss and fewer artifacts compared to the adversarial loss. (c) 2023 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license ( http://creativecommons.org/licenses/by-nc-nd/4.0/ )
资助项目	National Natural Science Foundation of China[62071466] ; Fund of National Key Laboratory of Science and Technology on Remote Sensing Information and Imagery Analysis, Beijing Research Institute of Uranium Geology[6142A010402] ; Guangxi Natural Science Foundation[2018GXNSFBA281086]
WOS关键词	QUALITY ASSESSMENT
WOS研究方向	Computer Science ; Engineering
语种	英语
出版者	ELSEVIER SCI LTD
WOS记录号	WOS:001009532700001
资助机构	National Natural Science Foundation of China ; Fund of National Key Laboratory of Science and Technology on Remote Sensing Information and Imagery Analysis, Beijing Research Institute of Uranium Geology ; Guangxi Natural Science Foundation
内容类型	期刊论文
源URL	[http://ir.ia.ac.cn/handle/173211/53513]
专题	多模态人工智能系统全国重点实验室
通讯作者	Huo, Chunlei
作者单位	1.Chinese Acad Sci, Inst Automat, Natl Lab Pattern Recognit, Beijing 100190, Peoples R China 2.Univ Chinese Acad Sci, Sch Artificial Intelligence, Beijing 100049, Peoples R China
推荐引用方式 GB/T 7714	An, Tai,Mao, Binjie,Xue, Bin,et al. Patch loss: A generic multi-scale perceptual loss for single image super-resolution[J]. PATTERN RECOGNITION,2023,139:12.
APA	An, Tai,Mao, Binjie,Xue, Bin,Huo, Chunlei,Xiang, Shiming,&Pan, Chunhong.(2023).Patch loss: A generic multi-scale perceptual loss for single image super-resolution.PATTERN RECOGNITION,139,12.
MLA	An, Tai,et al."Patch loss: A generic multi-scale perceptual loss for single image super-resolution".PATTERN RECOGNITION 139(2023):12.