FAST FUSION OF HYPERSPECTRAL AND MULTISPECTRAL IMAGES : A TUCKER APPROXIMATION APPROACH

Copyright (c) 2022 Clemence Prevost, Pierre Chainais, Remy Boyer
Contact: clemence.prevost@univ-lille.fr

This software reproduces the results from the following:

@unpublished{prevost:hal-03617759,
  TITLE = {{FAST FUSION OF HYPERSPECTRAL AND MULTISPECTRAL IMAGES : A TUCKER APPROXIMATION APPROACH}},
  AUTHOR = {Pr{\'e}vost, Cl{\'e}mence and Chainais, Pierre and Boyer, Remy},
  URL = {https://hal.archives-ouvertes.fr/hal-03617759},
  NOTE = {working paper or preprint},
  YEAR = {2022},
  MONTH = Mar,
  KEYWORDS = {hyperspectral super-resolution ; data fusion ; low-rank tensor factorizations ; recovery ; least-squares problem},
  PDF = {https://hal.archives-ouvertes.fr/hal-03617759/file/icip.pdf},
  HAL_ID = {hal-03617759},
  HAL_VERSION = {v1},
}

Link to the project

Content

• /metrics : contains the metrics used to assess the quality of the reconstruction
• /src : contains helpful files to run the demos

Minimal requirements

In order to run the demo files, you will need to:

• Download and install Tensorlab 3.0: https://www.tensorlab.net
• Download the dataset from the AVIRIS website: https://aviris.jpl.nasa.gov/dataportal/

Please quote the corresponding paper if you decide to use these codes.

How it works

This software reproduces the figures and tables contained in the paper. You can play with the two datasets.

Generate coupled tensor model

Real datasets are used. First, the HSI and MSI are generated following Wald's protocol. Then, white Gaussian noise is added to the observations.

Run algorithms

In fusion_isabella.m and fusion_lockwood.m, we showcase the performance of:

• The proposed algorithm, with [2,2] and [4,4] block-patterns
• SCOTT, for comparison

Plot results

The metrics and computation time are then displayed in a table. Slices of the reference and reconstructions are plotted in a figure. The color scale is generated according to the image's spectral support.

Available demos

They are available in the /demos folder. The table below summarized what does what

Name	Content
fusion_isabella.m	Simulations for Isabella Lake dataset
fusion_lockwood.m	Simulations for Lockwood dataset
choice_ranks.m	plots R-SNR as a function of the multilinear ranks