DR2-Net: Deep Residual Reconstruction Network for Image Compressive Sensing

By Hantao Yao, Feng Dai, Dongming Zhang, Yike Ma, Shiliang Zhang, Yongdong Zhang. Neurocomputing, 2017

352 citations recorded by the source corpus.

Abstract Most traditional algorithms for compressive sensing image reconstruction suffer from the intensive computation. Recently, deep learning-based reconstruction algorithms have been reported, which dramatically reduce the time complexity than iterative reconstruction algorithms. In this paper, we propose a novel Deep Residual Reconstruction Network (DR2-Net) to reconstruct the image from its Compressively Sensed (CS) measurement. The DR2-Net is proposed based on two observations: (1) linear mapping could reconstruct a high-quality preliminary image, and (2) residual learning could further improve the reconstruction quality. Accordingly, DR2-Net consists of two components, i.e., linear mapping network and residual network, respectively. Specifically, the fully-connected layer in neural network implements the linear mapping network. We then expand the linear mapping network to DR2-Net by adding several residual learning blocks to enhance the preliminary image. Extensive experiments demonstrate that the DR2-Net outperforms traditional iterative methods and recent deep learning-based methods by large margins at measurement rates 0.01, 0.04, 0.1, and 0.25, respectively. The code of DR2-Net has been released on: https://github.com/coldrainyht/caffe_dr2 .

Subfields: Computational Mechanics, Radiology, Nuclear Medicine and Imaging

View Paper: https://www.semanticscholar.org/paper/3b0c8925a39c4a1010ec5bbfac5cb653980c6f54

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Mechanics

In Radiology, Nuclear Medicine and Imaging

Citations in the corpus

Using Deep Neural Networks for Inverse Problems in Imaging: Beyond Analytical Methods (2018, methodology): “[27] use a similar deep residual CNN to reconstruct an image from its CS measurements.”
Deep speckle correlation: a deep learning approach toward scalable imaging through scattering media (2018, background): “DL is proven to be extremely powerful in solving complex imaging problems, providing performance surpassing those using state-of-the-art model-based techniques [34], such as super-resolution [43, 57], deconvolution [58, 61], compressive imaging [27, 60], tomography [19], holography [42,44], and phas”
Deep Learning for Massive MIMO CSI Feedback (2017, methodology): “Recently, several DL architectures have been proposed to reconstruct natural images from CS measurements [10]–[12].”
Convolutional Neural Network-Based Multiple-Rate Compressive Sensing for Massive MIMO CSI Feedback: Design, Simulation, and Analysis (2019, methodology): “From [19], it is a good strategy to generate a preliminary reconstruction at the decoder via a linear mapping network and then use a residual network to refine estimates.”
Learned D-AMP: Principled Neural Network based Compressive Image Recovery (2017, methodology): “Other papers in this line of work used either pure convolutional layers (DeepInverse [6]) or a combination of convolutional and fully-connected layers (DR(2)-Net [7] and ReconNet [5]) to build deep learning frameworks capable of solving the CS recovery problem.”