中文 | English

J. Hefei Univ. Tech. (Nat. Sci.)

合肥工业大学学报(自然科学版)

JOURNAL OF HEFEI UNIVERSITY OF TECHNOLOGY (NATURAL SCIENCE)

Competent Authorities: Ministry of Education of the People's Republic of China

Sponsor by: Hefei University of Technology

ISSN 1000-1646, CN 34-1001/N

Multi-scale image super-resolution reconstruction using channel attention

Download Full Text

注意力机制下的多尺度图像超分辨率重建

Journal Information

J. Hefei Univ. Tech. (Nat. Sci.), Sept.2024, Vol.47 No.9: 1255-1261

DOI: 10.3969/j.issn.1003-5060.2024.09.015

Authors

HE Qichen, HE Lei

(School of Mathematics, Hefei University of Technology, Hefei 230601, China)

Abstract and Keywords

Abstract: This paper combines the popular multi-scale convolution and channel attention mechanism, and proposes a novel convolutional neural network(CNN) structure, namely the multi-scale CNN under attention mechanism. A large number of residual structures are added to the proposed network structure, which deepens the depth of the network. The utilization of multi-scale convolution enables the network to extract richer information from pictures. The introduction of the attention mechanism enables the network to have greater weight in processing high-frequency information. Experimental results show that the multi-scale CNN under attention mechanism has achieved good performance in image super-resolution(SR) reconstruction, and the effect of image detail restoration is satisfactory.

Keywords: super-resolution (SR); deep learning; convolutional neural network (CNN); attention mechanism; multi-scale

作者信息

何启琛，何蕾

（合肥工业大学数学学院，安徽合肥230601）

摘要和关键词

摘要: 文章结合目前较流行的多尺度卷积和通道注意力机制，提出一种新颖的卷积神经网络（convolutional neural network，CNN）结构，即注意力机制下的多尺度卷积神经网络。该网络结构中加入大量的残差结构，加深了网络的深度；多尺度卷积的使用使该网络能从图片中提取更加丰富的信息；注意力机制的引入使网络处理高频信息时有更大的权重。实验结果表明，多尺度注意力机制卷积神经网络在图像超分辨率（super-resolution，SR）重建上取得了良好的表现，图像细节恢复效果令人满意。

关键词: 超分辨率(SR)；深度学习；卷积神经网络(CNN)；注意力机制；多尺度

Previous Next

Back to Contents