The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
Explorar as informações estruturais anteriores às imagens faciais é uma questão fundamental da super-resolução facial (SR). Embora as redes neurais convolucionais profundas (CNNs) possuam uma poderosa capacidade de representação, como usar com precisão as informações estruturais faciais continua sendo um desafio. Neste artigo, propusemos uma nova rede de fusão residual para utilizar a informação estrutural multiescala para a face SR. Diferente dos métodos existentes para aumentar a profundidade da rede, o módulo de atenção de gargalo é introduzido para extrair características estruturais faciais finas, explorando a correlação de mapas de características. Finalmente, escalas hierárquicas de informações estruturais são fundidas para gerar uma imagem facial de alta resolução (HR). Resultados experimentais mostram que a rede proposta supera alguns algoritmos de SR facial baseados em CNNs de última geração.
Yu WANG
Wuhan Institute of Technology
Tao LU
Wuhan Institute of Technology
Zhihao WU
Wuhan Institute of Technology
Yuntao WU
Wuhan Institute of Technology
Yanduo ZHANG
Wuhan Institute of Technology
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copiar
Yu WANG, Tao LU, Zhihao WU, Yuntao WU, Yanduo ZHANG, "Face Super-Resolution via Hierarchical Multi-Scale Residual Fusion Network" in IEICE TRANSACTIONS on Fundamentals,
vol. E104-A, no. 9, pp. 1365-1369, September 2021, doi: 10.1587/transfun.2020EAL2103.
Abstract: Exploring the structural information as prior to facial images is a key issue of face super-resolution (SR). Although deep convolutional neural networks (CNNs) own powerful representation ability, how to accurately use facial structural information remains challenges. In this paper, we proposed a new residual fusion network to utilize the multi-scale structural information for face SR. Different from the existing methods of increasing network depth, the bottleneck attention module is introduced to extract fine facial structural features by exploring correlation from feature maps. Finally, hierarchical scales of structural information is fused for generating a high-resolution (HR) facial image. Experimental results show the proposed network outperforms some existing state-of-the-art CNNs based face SR algorithms.
URL: https://global.ieice.org/en_transactions/fundamentals/10.1587/transfun.2020EAL2103/_p
Copiar
@ARTICLE{e104-a_9_1365,
author={Yu WANG, Tao LU, Zhihao WU, Yuntao WU, Yanduo ZHANG, },
journal={IEICE TRANSACTIONS on Fundamentals},
title={Face Super-Resolution via Hierarchical Multi-Scale Residual Fusion Network},
year={2021},
volume={E104-A},
number={9},
pages={1365-1369},
abstract={Exploring the structural information as prior to facial images is a key issue of face super-resolution (SR). Although deep convolutional neural networks (CNNs) own powerful representation ability, how to accurately use facial structural information remains challenges. In this paper, we proposed a new residual fusion network to utilize the multi-scale structural information for face SR. Different from the existing methods of increasing network depth, the bottleneck attention module is introduced to extract fine facial structural features by exploring correlation from feature maps. Finally, hierarchical scales of structural information is fused for generating a high-resolution (HR) facial image. Experimental results show the proposed network outperforms some existing state-of-the-art CNNs based face SR algorithms.},
keywords={},
doi={10.1587/transfun.2020EAL2103},
ISSN={1745-1337},
month={September},}
Copiar
TY - JOUR
TI - Face Super-Resolution via Hierarchical Multi-Scale Residual Fusion Network
T2 - IEICE TRANSACTIONS on Fundamentals
SP - 1365
EP - 1369
AU - Yu WANG
AU - Tao LU
AU - Zhihao WU
AU - Yuntao WU
AU - Yanduo ZHANG
PY - 2021
DO - 10.1587/transfun.2020EAL2103
JO - IEICE TRANSACTIONS on Fundamentals
SN - 1745-1337
VL - E104-A
IS - 9
JA - IEICE TRANSACTIONS on Fundamentals
Y1 - September 2021
AB - Exploring the structural information as prior to facial images is a key issue of face super-resolution (SR). Although deep convolutional neural networks (CNNs) own powerful representation ability, how to accurately use facial structural information remains challenges. In this paper, we proposed a new residual fusion network to utilize the multi-scale structural information for face SR. Different from the existing methods of increasing network depth, the bottleneck attention module is introduced to extract fine facial structural features by exploring correlation from feature maps. Finally, hierarchical scales of structural information is fused for generating a high-resolution (HR) facial image. Experimental results show the proposed network outperforms some existing state-of-the-art CNNs based face SR algorithms.
ER -