HyperQUEEN-MF: Hyperspectral Quantum Deep Network with Multi-Scale Feature Fusion For Quantum Image Super-Resolution

Hyperspectral images (HSIs) have drawn considerable attention over the last decade thanks to their high spectral resolution. However, as hardware limitations often cause the low spatial resolution of HSIs compared to traditional RGB and multispectral images, the hyperspectral image super-resolution...

Full description

Saved in:
Bibliographic Details
Published inProceedings of the IEEE Sensor Array and Multichannel Signal Processing Workshop pp. 1 - 5
Main Authors Hsu, Shih-Min, Lin, Tzu-Hsuan, Lin, Chia-Hsiang
Format Conference Proceeding
LanguageEnglish
Published IEEE 08.07.2024
Subjects
Feature extraction
hyperspectral image
hyperspectral super-resolution
Information retrieval
Inverse problems
quantum deep learning
quantum image processing
Quantum neural network
Satellites
Superresolution
Transformers
Transforms
Online AccessGet full text

Cover

More Information
Summary:Hyperspectral images (HSIs) have drawn considerable attention over the last decade thanks to their high spectral resolution. However, as hardware limitations often cause the low spatial resolution of HSIs compared to traditional RGB and multispectral images, the hyperspectral image super-resolution (HSR) problem becomes crucial in facilitating the subsequent material identification applications. Recently, we developed hyperspectral quantum deep network (HyperQUEEN) for satellite data restoration, thereby proving that quantum features do play some role in developing high-performance inverse imaging algorithms. Before HyperQUEEN, which elegantly performs the hyperspectral matrix/tensor factorization via the so-called low-rank module, other existing quantum image processing tasks can only achieve basic geometry transforms or simple classification-level tasks. In this work, we go a step further by investigating the quantum implementation of multi-scale feature fusion (MF) transformer and applying the HyperQUEEN-MF to solve the challenging inverse problem known as single HSR (SHSR). By learning multi-scale quantum features, the proposed HyperQUEEN-MF does show outstanding SHSR performances, while the experimental results also show the feasibility of quantum-driven hyperspectral super-resolution. This creates a door for solving other advanced challenging inverse problems via quantum feature information extraction and processing.
ISSN:2151-870X
DOI:10.1109/SAM60225.2024.10636692