A robust matching pursuit algorithm using information theoretic learning

Zhang, Miaohua; Gao, Yongsheng; Sun, Changming; Blumenstein, Michael

doi:10.1016/j.patcog.2020.107415

Cited by 5 publications

(1 citation statement)

References 39 publications

(58 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The typical algorithms using convex relaxation are Basis Pursuit (BP) [15] algorithm, Gradient Projection for Sparse Reconstruction (GPSR) [16] algorithm and Interior Point Method (IPM) [17] algorithm. Greedy algorithms [18,19] update the estimated signal support set iteratively to approximate the target signal, which includes two basic steps: atomic selection and signal updating estimation. Typical greedy algorithms mainly include Matching Pursuit (MP) [20] algorithm, Orthogonal Matching Pursuit (OMP) [21] algorithm and Compressive Sampling Matching Pursuit (CoSaMP) [19] algorithm.…”

Section: The Classical Framework Of Compressed Sensingmentioning

confidence: 99%

JSRNN: Joint Sampling and Reconstruction Neural Networks for High Quality Image Compressed Sensing

Zeng¹,

Ye²,

Wang³

et al. 2022

Preprint

View full text Add to dashboard Cite

Most Deep Learning (DL) based Compressed Sensing (DCS) algorithms adopt a single neural network for signal reconstruction, and fail to jointly consider the influences of the sampling operation for reconstruction. In this paper, we propose unified framework, which jointly considers the sampling and reconstruction process for image compressive sensing based on welldesigned cascade neural networks. Two sub-networks, which are the sampling sub-network and the reconstruction sub-network, are included in the proposed framework. In the sampling sub-network, an adaptive full connected layer instead of the traditional random matrix is used to mimic the sampling operator. In the reconstruction sub-network, a cascade network combining stacked denoising autoencoder (SDA) and convolutional neural network (CNN) is designed to reconstruct signals. The SDA is used to solve the signal mapping problem and the signals are initially reconstructed. Furthermore, CNN is used to fully recover the structure and texture features of the image to obtain better reconstruction performance. Extensive experiments show that this framework outperforms many other state-ofthe-art methods, especially at low sampling rates.

show abstract