基于通道注意力的多尺度全卷积压缩感知重构

doi:10.19678/j.issn.1000-3428.0063546

摘要/Abstract

摘要： 现有分块压缩感知图像重建算法存在计算量大、重构时间长，以及在低采样率下重构的图像具有严重的块效应等问题。提出一种基于通道注意力的多尺度全卷积压缩感知图像重构模型。通过均值滤波消除完整场景图像中的噪声点，减少冗余信息，以提取更加有效的信息，采用不同卷积核大小的卷积层对低频信息进行多尺度全卷积采样，得到不同感受野的图像特征信息，丰富网络中原始图像的特征信息。在此基础上，设计一种新的注意力残差模块，通过挖掘特征图通道之间的关联性以提取关键特征信息，提升重构图像的质量。在DIV2K、Set0和Set5数据集上的实验结果表明，当采样率为1%时，该模型的峰值信噪比和结构相似性相较于深度学习模型ISTA-Net分别平均提升了2.02 dB和0.078 2，相较于迭代优化模型TVAL3，重构一张256×256像素图像所花费的时间平均缩短2.608 4 s。所提模型在低采样率下能够有效利用原始图像中的信息生成更清晰的重构图像。

关键词: 压缩感知, 卷积神经网络, 均值滤波, 多尺度全卷积测量, 注意力机制

Abstract: Existing reconstruction algorithms for block compressive sensing images have problems, such as extensive computation, extended reconstruction time, and significant block effect of reconstructed images at low sampling rates.This study proposes a multi-scale, fully convolutional compressed sensing image reconstruction model based on channel attention.The complete scene image is subjected to mean filtering to eliminate noise points in the image, reduce redundant information, and receive more effective information.Convolution sampling is used to obtain image feature information for different receptive fields, enriching the feature information of the original image in the network.A new attention residual module is designed to extract critical feature information by mining the correlation between feature map channels to improve the reconstruction image quality.The experimental results on DIV2K, Set0, and Set5 show that when the sampling rate is 1%, the average Peak Signal-to-Noise Ratio(PSNR) and Structural Similarity(SSIM) are improved by 2.02 dB and 0.078 2, respectively, compared with the deep learning model, ISTA-Net.Compared with the iterative optimization model, TVAL3, the average time spent reconstructing a 256×256 pixel image is reduced by 2.608 4 s.The proposed model can effectively use the information in the original image to generate a clearer reconstructed image at a low sampling rate.

Key words: compressed sensing, Convolutional Neural Network(CNN), mean filtering, multi-scale fully convolutional measurement, attention mechanism

中图分类号:

TP391

刘玉红, 陈满银, 刘晓燕. 基于通道注意力的多尺度全卷积压缩感知重构[J]. 计算机工程, 2022, 48(12): 189-195.

LIU Yuhong, CHEN Manyin, LIU Xiaoyan. Multi-Scale Fully Convolutional Compressed Sensing Reconstruction Based on Channel Attention[J]. Computer Engineering, 2022, 48(12): 189-195.

https://www.ecice06.com/CN/Y2022/V48/I12/189

图/表 10

20230112183758

20230112183802

20230112183806

20230112183809

20230112183813

20230112183818

20230112183821

20230112183825

20230112183828

20230112183832

参考文献

[1] DONOHO D L.Compressed sensing[J].IEEE Transactions on Information Theory, 2006, 52(4):1289-1306.
[2] XIAO Y, YANG J, YUAN X.Alternating algorithms for total variation image reconstruction from random projections[J].Inverse Problems & Imaging, 2012, 6(3):547-563.
[3] DONG S W, SHI G M, LI X, et al.Image reconstruction with locally adaptive sparsity and nonlocal robust regularization[J].Signal Processing:Image Communication, 2012, 27(10):1109-1122.
[4] ELDAR Y C, KUPPINGER P, BOLCSKEI H.Block-sparse signals:uncertainty relations and efficient recovery[J].IEEE Transactions on Signal Processing, 2010, 58(6):3042-3054.
[5] KULKARNI K, LOHIT S, TURAGA P, et al.ReconNet:non-iterative reconstruction of images from compressively sensed measurements[C]//Proceedings of Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2016:449-458.
[6] HE K M, ZHANG X, REN S, et al.Deep residual learning for image recognition[C]//Proceedings of Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2016:770-778.
[7] YAO H, DAI F, ZHANG S, et al.DR²-Net:deep residual reconstruction network for image compressive sensing[J].Neurocomputing, 2019, 359:483-493.
[8] 刘玉红, 刘树英, 付福祥.基于卷积神经网络的压缩感知重构算法优化[J].计算机科学, 2020, 47(3):143-148. LIU Y H, LIU S Y, FU F X.Opimization of compressed sensing reconstruction algorithm optimization based on convolutional neural network[J].Computer Science, 2020, 47(3):143-148.(in Chinese)
[9] XIE X M, WANG Y X, SHI G M, et al.Adaptive measurement network for CS image reconstruction[C]//Proceedings of Chinese Conference on Computer Vision.Berlin, Germany:Springer, 2017:407-417.
[10] LIU J, ZHANG W, TANG Y, et al.Residual feature aggregation network for image super-resolution[C]//Proceedings of Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2020:2359-2368.
[11] MOU L, HUA Y, ZHU X X.Relation matters:relational context-aware fully convolutional network for semantic segmentation of high-resolution aerial images[J].IEEE Transactions on Geoscience and Remote Sensing, 2020, 58(11):7557-7569.
[12] ZHANG K, ZUO W, CHEN Y, et al.Beyond a Gaussian denoiser:residual learning of deep CNN for image denoising[J].IEEE Transactions on Image Processing, 2017, 26(7):3142-3155.
[13] MOUSAVI A, PATEL A B, BARANIUK R G.A deep learning approach to structured signal recovery[C]//Proceedings of the 53rd Annual Allerton Conference on Communication, Control, and Computing.Washington D.C., USA:IEEE Press, 2015:1336-1343.
[14] LI Y D, HAO Z, LEI H.Survey of convolutional neural network[J].Journal of Computer Applications, 2016, 36(9):2508-2515.
[15] 练秋生, 富利鹏, 陈书贞, 等.基于多尺度残差网络的压缩感知重构算法[J].自动化学报, 2019, 45(11):2082-2091. LIAN Q S, FU L P, CHEN S Z, et al.A compressed sensing algorithm based on multi-scale residual reconstruction network[J].Acta Automatica Sinica, 2019, 45(11):2082-2091.(in Chinese)
[16] ZHANG J, GHANEM B.ISTA-Net:interpretable optimization-inspired deep network for image compressive sensing[C]//Proceedings of Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2018:1828-1837.
[17] SUN Y, CHEN J, LIU Q, et al.Dual-path attention network for compressed sensing image reconstruction[J].IEEE Transactions on Image Processing, 2020, 29:9482-9495.
[18] DU J, XIE X, WANG C, et al.Perceptual compressive sensing[C]//Proceedings of Chinese Conference on Pattern Recognition and Computer Vision.Berlin, Germany:Springer, 2018:268-279.
[19] SHI W, JIANG F, ZHANG S, et al.Deep networks for compressed image sensing[C]//Proceedings of International Conference on Multimedia and Expo. Washington D.C., USA:IEEE Press, 2017:877-882.
[20] 杨春玲, 裴翰奇.基于残差学习的多阶段图像压缩感知神经网络[J].华南理工大学学报(自然科学版), 2020, 48(5):82-91. YANG C L, PEI H Q.Multistage image compressive sensing neural network based on residual learning[J].Journal of South China University of Technology (Natural Science Edition), 2020, 48(5):82-91.(in Chinese).
[21] DU J, XIE X, WANG C, et al.Fully convolutional measurement network for compressive sensing image reconstruction[J].Neurocomputing, 2019, 328:105-112
[22] ZHANG Z, GAO D, XIE X, et al.Dual-channel reconstruction network for image compressive sensing[J].Sensors, 2019, 19(11):2549.
[23] TAO X, HONG X, SHI W, et al.Analogy-detail networks for object recognition[J].IEEE Transactions on Neural Networks and Learning Systems, 2020, 32(10):4404-4418.
[24] WANG F, TAX D M J.Survey on the attention based RNN model and its applications in computer vision[EB/OL].[2021-11-19].https://arxiv.org/abs/1601.06823.pdf.
[25] HU J, SHEN L, ALBANIE S, et al.Squeeze-and-excitation networks[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(8):2011-2023.
[26] WOO S, PARK J, LEE J Y, et al.CBAM:convolutional block attention module[C]//Proceedings of European Conference on Computer Vision.Berlin, Germany:Springer, 2018:3-19.
[27] LI C, YIN W, JIANG H, et al.An efficient augmented Lagrangian method with applications to total variation minimization[J].Computational Optimization and Applications, 2013, 56(3):507-530.

选择文件类型/文献管理软件名称

选择包含的内容