An image compressed sensing algorithm based on adaptive nonlinear network

doi:10.1088/1674-1056/ab7b4e

中国物理B ›› 2020, Vol. 29 ›› Issue (5): 54203-054203.doi: 10.1088/1674-1056/ab7b4e

• ELECTROMAGNETISM, OPTICS, ACOUSTICS, HEAT TRANSFER, CLASSICAL MECHANICS, AND FLUID DYNAMICS • 上一篇下一篇

An image compressed sensing algorithm based on adaptive nonlinear network

Yuan Guo(郭媛), Wei Chen(陈炜), Shi-Wei Jing(敬世伟)

School of Computer and Control Engineering, Qiqihar University, Qiqihar 161006, China

收稿日期:2019-11-01 修回日期:2020-02-01 出版日期:2020-05-05 发布日期:2020-05-05
通讯作者: Wei Chen E-mail:1010172469@qq.com
基金资助:
Project supported by the National Natural Science Foundation of China (Grant No. 61872204), the Natural Science Fund of Heilongjiang Province, China (Grant No. F2017029), the Scientific Research Project of Heilongjiang Provincial Universities, China (Grant No. 135109236), and the Graduate Research Project, China (Grant No. YJSCX2019042).

An image compressed sensing algorithm based on adaptive nonlinear network

Yuan Guo(郭媛), Wei Chen(陈炜), Shi-Wei Jing(敬世伟)

School of Computer and Control Engineering, Qiqihar University, Qiqihar 161006, China

Received:2019-11-01 Revised:2020-02-01 Online:2020-05-05 Published:2020-05-05
Contact: Wei Chen E-mail:1010172469@qq.com
Supported by:
Project supported by the National Natural Science Foundation of China (Grant No. 61872204), the Natural Science Fund of Heilongjiang Province, China (Grant No. F2017029), the Scientific Research Project of Heilongjiang Provincial Universities, China (Grant No. 135109236), and the Graduate Research Project, China (Grant No. YJSCX2019042).

摘要/Abstract

摘要： Traditional compressed sensing algorithm is used to reconstruct images by iteratively optimizing a small number of measured values. The computation is complex and the reconstruction time is long. The deep learning-based compressed sensing algorithm can greatly shorten the reconstruction time, but the algorithm emphasis is placed on reconstructing the network part mostly. The random measurement matrix cannot measure the image features well, which leads the reconstructed image quality to be improved limitedly. Two kinds of networks are proposed for solving this problem. The first one is ReconNet's improved network IReconNet, which replaces the traditional linear random measurement matrix with an adaptive nonlinear measurement network. The reconstruction quality and anti-noise performance are greatly improved. Because the measured values extracted by the measurement network also retain the characteristics of image spatial information, the image is reconstructed by bilinear interpolation algorithm (Bilinear) and dilate convolution. Therefore a second network USDCNN is proposed. On the BSD500 dataset, the sampling rates are 0.25, 0.10, 0.04, and 0.01, the average peak signal-noise ratio (PSNR) of USDCNN is 1.62 dB, 1.31 dB, 1.47 dB, and 1.95 dB higher than that of MSRNet. Experiments show the average reconstruction time of USDCNN is 0.2705 s, 0.3671 s, 0.3602 s, and 0.3929 s faster than that of ReconNet. Moreover, there is also a great advantage in anti-noise performance.

关键词: compressed sensing, deep learning, bilinear interpolation, dilate convolution

Abstract: Traditional compressed sensing algorithm is used to reconstruct images by iteratively optimizing a small number of measured values. The computation is complex and the reconstruction time is long. The deep learning-based compressed sensing algorithm can greatly shorten the reconstruction time, but the algorithm emphasis is placed on reconstructing the network part mostly. The random measurement matrix cannot measure the image features well, which leads the reconstructed image quality to be improved limitedly. Two kinds of networks are proposed for solving this problem. The first one is ReconNet's improved network IReconNet, which replaces the traditional linear random measurement matrix with an adaptive nonlinear measurement network. The reconstruction quality and anti-noise performance are greatly improved. Because the measured values extracted by the measurement network also retain the characteristics of image spatial information, the image is reconstructed by bilinear interpolation algorithm (Bilinear) and dilate convolution. Therefore a second network USDCNN is proposed. On the BSD500 dataset, the sampling rates are 0.25, 0.10, 0.04, and 0.01, the average peak signal-noise ratio (PSNR) of USDCNN is 1.62 dB, 1.31 dB, 1.47 dB, and 1.95 dB higher than that of MSRNet. Experiments show the average reconstruction time of USDCNN is 0.2705 s, 0.3671 s, 0.3602 s, and 0.3929 s faster than that of ReconNet. Moreover, there is also a great advantage in anti-noise performance.

Key words: compressed sensing, deep learning, bilinear interpolation, dilate convolution

中图分类号: (Image reconstruction; tomography)

42.30.Wb

42.68.Sq (Image transmission and formation)

郭媛, 陈炜, 敬世伟. An image compressed sensing algorithm based on adaptive nonlinear network[J]. 中国物理B, 2020, 29(5): 54203-054203.

Yuan Guo(郭媛), Wei Chen(陈炜), Shi-Wei Jing(敬世伟). An image compressed sensing algorithm based on adaptive nonlinear network[J]. Chin. Phys. B, 2020, 29(5): 54203-054203.

参考文献 30

[1]	Donoho D L 2006 IEEE Trans. Inf. Theory 52 1289 doi: 10.1109/TIT.2006.871582
[2]	Candes E J, Romberg J and Tao T 2006 IEEE Trans. Inf. Theory 52 489 doi: 10.1109/TIT.2005.862083
[3]	Candes E J and Wakin M B 2008 IEEE Signal Process. Mag. 25 21 doi: 10.1109/MSP.2007.914731
[4]	Rani M, Dhok S B and Deshmukh R B 2018 IEEE Access 6 4875 doi: 10.1109/ACCESS.2018.2793851
[5]	Liu X J, Xia S T and Fu F W 2017 IEEE Trans. Inf. Theory 63 2922
[6]	Moshtaghpour A, Jacques L, Cambareri V, Degraux K and Vleeschouwer C D 2016 IEEE Signal Process. Lett. 23 25 doi: 10.1109/LSP.2015.2497543
[7]	Nguyen N, Needell D and Woolf T 2017 IEEE Trans. Inf. Theory 63 6869 doi: 10.1109/TIT.2017.2749330
[8]	Lee J, Choi J W and Shim B 2016 J. Commun. Netw. 18 699 doi: 10.1109/JCN.2016.000100
[9]	Wang R, Zhang J L, Ren S L and Li Q J 2016 Tsinghua Sci. Technol. 21 71 doi: 10.1109/TST.2016.7399284
[10]	Davenport M A, Needell D and Wakin M B 2013 IEEE Trans. Inf. Theory 59 6820 doi: 10.1109/TIT.2013.2273491
[11]	Kong M, Chen M S, zhang L, Cao X Y and Wu X L 2016 Chin. Phys. Lett. 33 018402 doi: 10.1088/0256-307X/33/1/018402
[12]	Zhang Y X, Li Y Z, Wang Z Y, Song Z H, Lin R, Qian J Q, Yao J N 2019 Meas. Sci. Technol. 30 025402 doi: 10.1088/1361-6501/aaf4e7
[13]	Zhao R Q, Wang Q, Fu J and Ren L Q 2020 IEEE Trans. Image Process. 29 1654 doi: 10.1109/TIP.2019.2944722
[14]	Mousavi A, Patel A B and Baraniuk R G 2015 53rd Annual Allerton Conference on Communication, Control, and Computing, September 29-October 2, 2015, Monticello, USA, p. 1336
[15]	Kulkarni K, Lohit S, Turaga P, Kerviche R and Ashok A 2016 IEEE Conference on Computer Vision and Pattern Recognition, June 27-30, 2016, Las Vegas, USA, p. 449
[16]	Dong C, Loy C C, He K and Tang X 2016 IEEE Trans. Pattern Anal. Mach. Intell. 38 295 doi: 10.1109/TPAMI.2015.2439281
[17]	Yao H T, Dai F, Zhang SL, Zhang Y D, Tian Q and Xu C S 2019 Neurocomputing 359 483 doi: 10.1016/j.neucom.2019.05.006
[18]	He K M, Zhang X Y, Ren S Q and Sun J 2016 IEEE Conference on Computer Vision and Pattern Recognition, June 27-30, 2016, Las Vegas, USA, p. 770
[19]	Lian Q S, Fu L P, Chen S Z and Shi B S 2019 Acta Autom. Sin. 45 2082
[20]	Zhang Z D, Wang X R and Jung C 2019 IEEE Trans. Image Process. 28 1625 doi: 10.1109/TIP.2018.2877483
[21]	Ioffe S and Szegedy C 2015 Proceedings of the 32nd International Conference on Machine Learning, July 6-11, 2015, Lille, France, p. 448
[22]	Liu Y N, Niu H Q and Li Z L 2019 Chin. Phys. Lett. 36 044302 doi: 10.1088/0256-307X/36/4/044302
[23]	Kingma D P and Ba J L 2014 arXiv: 1412.6980v9 [cs.LG]
[24]	Shi W Z, Caballero J, Huszár F, Totz J, Aitken A P, Bishop R, Rueckert D and Wang Z H 2016 IEEE Conference on Computer Vision and Pattern Recognition, June 27-30, 2016, Las Vegas, USA, p. 1874
[25]	Zeiler M D, Taylor G W and Fergus R 2019 Acta Phys. Sin. 68 200501 (in Chinese)
[26]	Li C B, Yin W T, Jiang H and Zhang Y 2013 Comput. Optim. Appl. 56 507 doi: 10.1007/s10589-013-9576-1
[27]	Dong W S, Shi G M, Li X, Ma Y and Huang F 2014 IEEE Trans. Image Process. 23 3618 doi: 10.1109/TIP.2014.2329449
[28]	Metzler C A, Maleki A and Baraniuk R G 2016 IEEE Trans. Inf. Theory 62 5117 doi: 10.1109/TIT.2016.2556683
[29]	Dabov K, Foi A, Katkovnik V and Egiazarian K 2007 IEEE Trans. Image Process. 16 2080 doi: 10.1109/TIP.2007.901238
[30]	Shi H and Wang L D 2019 Acta Phys. Sin. 68 200501 (in Chinese)

An image compressed sensing algorithm based on adaptive nonlinear network

An image compressed sensing algorithm based on adaptive nonlinear network

RichHTML

PDF (PC)

赞

可视化

摘要/Abstract

引用本文

使用本文

参考文献 30

相关文章 15

Metrics

本文评价

推荐阅读 0

[1]	Xiao-Gang Wang(汪小刚) and Hao-Yu Wei(魏浩宇). Deep-learning-based cryptanalysis of two types of nonlinear optical cryptosystems[J]. 中国物理B, 2022, 31(9): 94202-094202.
[2]	Xun Guo(郭寻), Hao Wang(王浩), Changkai Li(李长楷),Shijun Zhao(赵仕俊), Ke Jin(靳柯), and Jianming Xue(薛建明). Development of an electronic stopping power model based on deep learning and its application in ion range prediction[J]. 中国物理B, 2022, 31(7): 73402-073402.
[3]	Li-Jun Chang(常莉君), Yi-Fan Mo(莫一凡), Li-Ming Ling(凌黎明), and De-Lu Zeng(曾德炉). Data-driven parity-time-symmetric vector rogue wave solutions of multi-component nonlinear Schrödinger equation[J]. 中国物理B, 2022, 31(6): 60201-060201.
[4]	Gaoyi Lei(雷高益), Chencheng Tang(唐陈成), and Yueyang Zhai(翟跃阳). Fringe removal algorithms for atomic absorption images: A survey[J]. 中国物理B, 2022, 31(5): 50313-050313.
[5]	Xun-Qin Huo(火勋琴), Wei-Feng Yang(杨玮枫), Wen-Hui Dong(董文卉), Fa-Cheng Jin(金发成), Xi-Wang Liu(刘希望), Hong-Dan Zhang(张宏丹), and Xiao-Hong Song(宋晓红). Review on typical applications and computational optimizations based on semiclassical methods in strong-field physics[J]. 中国物理B, 2022, 31(3): 33101-033101.
[6]	Qiwen Xu(徐启文), Zhu Zheng(郑铸), and Huabei Jiang(蒋华北). Deep learning for image reconstruction in thermoacoustic tomography[J]. 中国物理B, 2022, 31(2): 24302-024302.
[7]	Shao-Ying Meng(孟少英), Mei-Yi Chen(陈美伊), Jie Ji(季杰), Wei-Wei Shi(史伟伟), Qiang Fu(付强), Qian-Qian Bao(鲍倩倩), Xi-Hao Chen(陈希浩), and Ling-An Wu(吴令安). Iterative filtered ghost imaging[J]. 中国物理B, 2022, 31(2): 28702-028702.
[8]	Tian-Shou Liang(梁添寿), Peng-Peng Shi(时朋朋), San-Qing Su(苏三庆), and Zhi Zeng(曾志). Learning physical states of bulk crystalline materials from atomic trajectories in molecular dynamics simulation[J]. 中国物理B, 2022, 31(12): 126402-126402.
[9]	Chengwei Deng(邓成伟), Yunxin Tang(唐蕴芯), Jian Zhang(张建), Wenfei Li(李文飞), Jun Wang(王骏), and Wei Wang(王炜). RNAGCN: RNA tertiary structure assessment with a graph convolutional network[J]. 中国物理B, 2022, 31(11): 118702-118702.
[10]	Jun-Cai Pu(蒲俊才), Jun Li(李军), and Yong Chen(陈勇). Soliton, breather, and rogue wave solutions for solving the nonlinear Schrödinger equation using a deep learning method with physical constraints[J]. 中国物理B, 2021, 30(6): 60202-060202.
[11]	Yi-Yi Huang(黄祎祎), Chen Ou-Yang(欧阳琛), Ke Fang(方可), Yu-Feng Dong(董玉峰), Jie Zhang(张杰), Li-Ming Chen(陈黎明), and Ling-An Wu(吴令安). High speed ghost imaging based on a heuristic algorithm and deep learning[J]. 中国物理B, 2021, 30(6): 64202-064202.
[12]	Wanrun Jiang(姜万润), Yuzhi Zhang(张与之), Linfeng Zhang(张林峰), and Han Wang(王涵). Accurate Deep Potential model for the Al-Cu-Mg alloy in the full concentration space[J]. 中国物理B, 2021, 30(5): 50706-050706.
[13]	Xing He(何行), Sheng-Mei Zhao(赵生妹), and Le Wang(王乐). Handwritten digit recognition based on ghost imaging with deep learning[J]. 中国物理B, 2021, 30(5): 54201-054201.
[14]	Hao Zhang(张浩), Yunjie Xia(夏云杰), and Deyang Duan(段德洋). Computational ghost imaging with deep compressed sensing[J]. 中国物理B, 2021, 30(12): 124209-124209.
[15]	王乐, 赵生妹. Compressed ghost imaging based on differential speckle patterns[J]. 中国物理B, 2020, 29(2): 24204-024204.