Deep Compressive Sensing Image Reconstruction NetworkBased on Non-Local Prior br

Journal of Electronics & Information Technology(2023)

Cited 0|Views3
No score
Abstract
The traditional iterative-based Compressive Sensing (CS) image reconstruction algorithm is easy tointegrate image prior information, but it has shortcomings such as insufficient performance and highcomputational complexity. The performance of the image reconstruction algorithm based on deep learning isbetter than the traditional reconstruction algorithm significantly, and it has lower time cost. Therefore, in orderto design a deep learning image reconstruction algorithm that uses prior information more effectively, a deepcompressive sensing image reconstruction network based on non-local priors is proposed. Firstly, the sparsenessand non-local prior are combined to establish a compressed sensing image reconstruction model. Secondly, the modelis decomposed into three sub-problems by the half quadratic splitting method. The solution of each sub-problemis carried out under the framework of deep learning. Finally, an end-to-end trainable image reconstructionmodel is jointly established. Simulation experiments show that the peak signal-to-noise ratio of the proposedalgorithm under the tested sampling rate and dataset is improved by 0.18 dB, 1.59 dB, 2.09 dB on averagecompared with the current mainstream reconstruction algorithm SCSNet, CSNet, ISTA-Net+ respectively.
More
Translated text
Key words
Compressive sensing,Image reconstruction,Deep learning,Non-local prior,Half quadratic splitting
AI Read Science
Must-Reading Tree
Example
Generate MRT to find the research sequence of this paper
Chat Paper
Summary is being generated by the instructions you defined