A Comparison Between Fourier and Hadamard Single-Pixel Imaging in Deep Learning-Enhanced Image Reconstruction

Many Single Pixel Imaging (SPI) schemes exist to reconstruct images, where the most notable schemes are Hadamard SPI (HSPI) and Fourier SPI (FSPI) effectively. To date, there exist comparisons between both methods, but only within the conventional optical image processing setting. With recent advanc...

Full description

Saved in:
Bibliographic Details
Published in:IEEE sensors letters 2023-09, Vol.7 (9), p.1-4
Main Authors: Lim, JiaYou, Roslan, Muhammad Razin, Lim, JunYi, MonnBaskaran, Vishnu, Chiew, YeongShiong, Phan, Raphael C.-W., Wang, Xin
Format: Article
Language:English
Subjects:
Compressed sensing
Compressive Sensing
Deep learning
Image enhancement
Image processing
Image reconstruction
Image restoration
Imaging
Indexing
Pixels
Real-time systems
Signal to noise ratio
Single-Pixel Imaging
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Many Single Pixel Imaging (SPI) schemes exist to reconstruct images, where the most notable schemes are Hadamard SPI (HSPI) and Fourier SPI (FSPI) effectively. To date, there exist comparisons between both methods, but only within the conventional optical image processing setting. With recent advancements in deep learning (DL), image restoration models exhibit considerable performance that could potentially be reformulated to enhance existing SPI schemes. In this work, we present the first-known comparison of conventional HSPI, FSPI, and their DL-enhanced variants, based on state-of-the-art NAFNet. The experiments are conducted by reconstructing images of the STL-10 dataset, followed by evaluations on the Set11, Set14, BSD68 and Urban100 test sets. Our experimental results show that DL-enhanced FSPI and HSPI achieved substantial performance gains on PSNR and SSIM. Interestingly, the improvement trend in PSNR for FSPI is inconsistent with HSPI due to the reconstructed graphical artifacts at higher sampling rates.
ISSN:2475-1472
2475-1472
DOI:10.1109/LSENS.2023.3303046