An 81.92Gpixels/s Fast Reconstruction of Images from Compressively Sensed Measurements

Corresponding author
1Hosei University
ISCAS 2022

Abstract

The reconstruction of Compressed Sensing is iteration-based and contains numerous divisions, thereby costing tremendous processing time.
To eliminate divisions, we adopt a sparse sensing matrix consisting mainly of zero-vectors. After removing zero-vectors, an invertible full-rank matrix is obtained. This allows the iteration-based reconstruction procedure to be replaced by a single matrix multiplication.

The proposed architecture is verified on the Xilinx Artix-7 FPGA. The result shows that our work accelerates the state-of-the-art method by 65 × and achieves 81.92Gpixels/s reconstruction.

Motivation

DDNeRF_Architecture_v21

Top: The architecture of PGTformer.

Methods

DDNeRF_Architecture_v21

Top: The architecture of PGTformer.

Experimental Results

DDNeRF_Architecture_v21

Quantitative comparison on VFHQ Blind setting.

BibTeX

@INPROCEEDINGS{9937930,
  author={Xu, Jiayao and Chi, Pham Do Kim and Fu, Chen and Zhou, Jinjia},
  booktitle={2022 IEEE International Symposium on Circuits and Systems (ISCAS)}, 
  title={An 81.92Gpixels/s Fast Reconstruction of Images from Compressively Sensed Measurements}, 
  year={2022},
  volume={},
  number={},
  pages={2978-2982},
  keywords={Costs;Reconstruction algorithms;Hardware;Matrices;Time measurement;Sensors;Sparse matrices;Compressed Sensing;Reconstruction Algorithm;Real-time Reconstruction;Hardware implementation},
  doi={10.1109/ISCAS48785.2022.9937930}}