MADNet: A Fast and Lightweight Network for Single-Image Super Resolution

作     者：Lan, Rushi Sun, Long Liu, Zhenbing Lu, Huimin Pang, Cheng Luo, Xiaonan 

作者机构：Guilin Univ Elect Technol Guangxi Key Lab Image & Graph Intelligent Proc Guilin 541004 Peoples R China South China Univ Technol Sch Comp Sci & Engn Guangzhou 510006 Peoples R China Kyushu Inst Technol Dept Mech & Control Engn Kitakyushu Fukuoka 8048550 Japan Guilin Univ Elect Technol Natl Local Joint Engn Res Ctr Satellite Nav & Loc Guilin 541004 Peoples R China 

出 版 物：《IEEE TRANSACTIONS ON CYBERNETICS》 (IEEE Trans. Cybern.)

年 卷 期：2021年第51卷第3期

页      面：1443-1453页

核心收录：

学科分类：0808[工学-电气工程] 08[工学] 0811[工学-控制科学与工程] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：National Natural Science Foundation of China [61702129, 61772149, U1701267, 61866009] National Key Research and Development Program of China [2018AAA0100305] China Postdoctoral Science Foundation [2018M633047] Guangxi Science and Technology Project [2019GXNSFAA245014, AD18281079, AD18216004, 2017GXNFDA198025, AA18118039] Innovation Project of GUET Graduate Education [2019YCXS048] 

主　　题：Feature extraction Computational modeling Image reconstruction Image resolution Task analysis Computer architecture Cybernetics Channel attention dense connections image super resolution lightweight multiscale mechanism 

摘      要：Recently, deep convolutional neural networks (CNNs) have been successfully applied to the single-image super-resolution (SISR) task with great improvement in terms of both peak signal-to-noise ratio (PSNR) and structural similarity (SSIM). However, most of the existing CNN-based SR models require high computing power, which considerably limits their real-world applications. In addition, most CNN-based methods rarely explore the intermediate features that are helpful for final image recovery. To address these issues, in this article, we propose a dense lightweight network, called MADNet, for stronger multiscale feature expression and feature correlation learning. Specifically, a residual multiscale module with an attention mechanism (RMAM) is developed to enhance the informative multiscale feature representation ability. Furthermore, we present a dual residual-path block (DRPB) that utilizes the hierarchical features from original low-resolution images. To take advantage of the multilevel features, dense connections are employed among blocks. The comparative results demonstrate the superior performance of our MADNet model while employing considerably fewer multiadds and parameters.