DR2-Net: Deep Residual Reconstruction Network for image compressive sensing

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 (DR 2 -Net) to reconstruct the image from its Compressively Sensed (CS) measurement. The DR 2 -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, DR 2 -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 DR 2 -Net by adding several residual learning blocks to enhance the preliminary image. Extensive experiments demonstrate that the DR 2 -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 DR 2 -Net has been released on: https://github.com/coldrainyht/caffe dr2 33 33 33 33 33 33 64 32 1 Block CS Measurements Linear Mapping Residual Learning Block

Convolutional Neural Network-Based Multiple-Rate Compressive Sensing for Massive MIMO CSI Feedback: Design, Simulation, and Analysis

“…In [17], [34], the RefineNet block is directly used as the last layer. However, in CsiNet, a convolutional layer follows the last RefineNet block, thereby disturbing the refinement.…”

“…Training the entire neural network DR 2 -Net in [17] is conducted in two steps. First, the encoder and the first FC layer at the decoder are trained using a large learning rate to obtain a preliminary reconstructed image.…”

“…From [17], 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. In [18], the random Gaussian measurement matrix is replaced by a learned measurement matrix at the encoder.…”

Data-Driven Seismic Waveform Inversion: A Study on the Robustness and Generalization

“…We first provide relevant literatures in solving inverse problems from other domains. In particular, we focus on deep neural network related techniques [1,14,19,49,50,55]. In general, those different deep-learning based methods for solving inverse problems can be categorized into four types [29]: 1) to learn an end-to-end regression with vanilla convolutional neural network (CNN), 2) to learn higher-level representation, 3) to gradual refinement of inversion procedure, and 4) to incorporate with analytical methods and to learn a denoiser.…”

Multi-Scale Deep Compressive Sensing Network

“…Since the linear projection of CS represented by = Φ can be understood as a fully connected layer having an identity activation and without bias [11], the fully connected layer shares the same practical problem of high signal dimensionality as the conventional frame-based compressed sensing. As a result, most research on DL-based CS (DCS) has focused on block-based schemes [11][12][13][14] and relied mostly on the learned prior from big data. Recently, multi-scale prior has been applied in image reconstruction as can be seen in the layer-wise wavelet network in [15], or multiwavelet convolutional network (MWCNN) [10] which greatly improves the restoration performance.…”