Matlab code for fast determination of eigenvalues of multilook polarimetric SAR data in the covariance matrix representation and for establishing the Loewner order of such matrices is given (in a zip file) with the papers
The Loewner Order and Direction of Detected Change in Sentinel-1 and Radarsat-2 Data (which describes the methods)
Fast matrix based computation of eigenvalues and the Loewner order in PolSAR data (which describes the fast implementations in the software).
If you use the code given here you must cite either of or both these papers.
Matlab code to perform change detection in a time series of multilook polarimetric SAR data in the covariance matrix representation is given (in a zip file) with the papers
Determining the points of change in time series of polarimetric SAR data (which describes the method)
Visualization of and software for omnibus test based change detected in a time series of polarimetric SAR data (which describes visualizations of change detected and software).
Such data may be obtained from spaceborne instruments such as ALOS, COSMO-SkyMed, RADARSAT-2, Sentinel-1, TerraSAR-X, or Yaogan.
If you use the code given here or
Dr. Morton J. Canty's
ENVI/IDL code or his
Docker/Google Earth Engine versions, you must cite either of or both these papers.
Matlab code to perform change detection between multilook polarimetric SAR data in the covariance matrix representation acquired at two time points, is given (in a zip file) with the paper
Change Detection in Full and Dual Polarization, Single- and Multi-Frequency SAR Data.
If you use this code you must cite this paper.
Matlab code to calculate kernel versions of
principal component analysis (PCA),
maximum autocorrelation factor (MAF)
and kernel minimum noise fraction (MNF) analysis
is given (in a zip file) with
Kernel maximum autocorrelation factor and minimum noise fraction transformations.
The code supports ENVI or ENVI-like header files.
If you use this software you must cite this paper.
Matlab code to perform multivariate alteration detection
maximum autocorrelation factor (MAF) analysis,
canonical correlation analysis (CCA) and
principal component analysis (PCA)
on multivariate image data
can be obtained here.
Versions supporting ENVI or ENVI-like header files
including code for the iteratively reweighted (IR-MAD) method
and associated automatic normalization,
are available also
method was developed by
our original MAD paper
(with James J. Simpson, University of California San Diego)
my IR-MAD paper
on an iterated extension to the original method.
Come back and check for new versions from time to time
<!(last update 11 Apr 2007;>
<!(last update 5 Nov 2008;>
<!(last update 19 Sep 2010;>
(last update 20 Sep 2010;
code for non-header versions is not updated anymore).
If you use this software do not forget to acknowledge the source.
If you use the function(s) for IR-MAD (also called iMAD) analysis
you must cite my
if you use the function(s) for IR-MAD/iMAD normalization you must cite the
IR-MAD normalization paper
by Morton J. Canty and myself.
especially on blunders in the code are most welcome.
Dr. Morton J. Canty
has written several extensions for the
remote sensing environment in
kernel PCA, the kernel MAF/MNF transformations,
IR-MAD change detection, automatic radiometric normalization using
MAD, and change detection in time series of covariance matrix multilook polSAR data.
The software is freely available and is described in his
Image Analysis, Classification and Change Detection in Remote Sensing: With Algorithms for ENVI/IDL and Python,
fourth revised edition, Taylor & Francis, CRC Press, 2019.
Some of my newer code was written partly within
DataBio, see also
Data-Driven Bioeconomy (2017-2019) under the Information and Communication Technologies Call of the EU Horizon 2020 Programme.
Some of both Mort's and my older code was written partly within
Global MOnitoring for Stability and Security (2004-2008),
a Network of Excellence
Aeronautics and Space
Priority of the Sixth Framework Programme
of the European Union.
Here are some other programs:
I have written a few computer
programs myself, and I have initiated and/or influenced work resulting in
a number of programs written by colleagues, Ph.D. students and M.Sc. students.
These programs center around statistics,
multivariate analysis, spatial (geostatistical)
data analysis, and hyper-spectral (remote sensing) tools.
In several programs the data may be sampled on a regular grid as well as
The most important ones are listed alphabetically below
(with co-workers/program authors mentioned).
- `rectangular' ASCII data to IMM defined irregular HIPS format
- band-interleave by line to IMM defined HIPS BIL format
- bioopt - bio-optical modelling by the matrix inversion method
- band-interleave by pixel to IMM defined HIPS BIP format
- boxcox - calculate Box-Cox transformation
- cancorr - canonical correlations analysis
- constrained energy minimisation (CEM), also known as matched filtering
- chi2 - read chi square test statistic and output significance level
- cloude - Cloude/Pottier decomposition of complex polarimetric radar signal
- cokrig - 2-D cokriging estimation of irregularly sampled data
- <!a href="http://www.imm.dtu.dk/~aa/cov2corr.pdf">
- calculate correlation matrix from variance/covariance matrix
- contingency table (or crosstabulation) analysis
- crossv - calculate traditional 1- and 2-D cross-semivariograms,
cross-covariance and cova functions of irregularly spaced data
- crossv2d - calculate various types of (1- and) 2-D cross-semivariograms,
cross-covariance and cova functions of irregularly spaced data
- decorr - RGB to principal components, stretch PCs, PCs to RGB
- pixelwise, hierarchical and contextual classification with feature selection
- distdisp - calculate distance between two variance/covariance matrices
- Wishart test for equal dispersion matrices
- Wishart based test for simultaneous equal means and dispersion matrices
- fuzzy c-means spectral and spatial cluster analysis
(Klaus Baggesen Hilger
- gamv2h - HIPS driver for calculation of cross-semivariograms and
cross-covariance functions with GSLIB's gamv2
<!li get_train - get observations under mask (to stdout column-wise)>
- histogram match to beta distribution
- hpca - Hebbian (linear) principal component analysis
- icda2.m - iterated canonical discriminant analysis for two groups
- transform a 3-frame sequence from IHS to RGB
- ihsdecorr - RGB to IHS, stretch S, IHS to RGB
- imaging - statistical image analysis for Microsoft Windows
(Johan Doré Hansen
- imshowrgb.m - display three bands from image cube as RGB with good stretching
<!isodata - unsupervised clustering>
- kcca - kernel canonical correlation analysis (CCA)
- kcem - kernel constrained energy minimization (CEM)
- kernel maximum autocorrelation factor (MAF) analysis
- kernel minimum noise fraction (MNF) analysis
- kernel principal component analysis (PCA)
- ktcimf - kernel target constrained interference minimization filter (TCIMF)
- 2-D (simple, ordinary or universal) kriging estimation of irregularly
spaced data (Henrik Juul Hansen)
- ktb3dh - HIPS driver for 2- or 3-D (simple, ordinary or universal) kriging
estimation of irregularly spaced data with GSLIB's ktb3d
- make location map of IMM defined irregular HIPS data
- logres - logarithmic and supremum residuals
- perform a wide range of multivariate orthogonal transformations such
as principal component, MAF, MNF, canonical analyses, etc.
- musecc - multiset canonical correlations analysis
- replace value with difference with nearest neighbour in irregularly spaced data
- orthogonal subspace projection
- one-way analysis of variance
- <!a href=http://www.imm.dtu.dk:/~aa/pls.pdf">pls
- partial least squares (PLS) regression
- project - project data in hyper-dimensional feature space onto a vector
- regularise class dispersion matrices
- transform multiples of three frames from RGB to IHS
- calculate linear and rank correlations
- robust principal components analysis
- saturate, standardize and stretch linearly
- simplestats - simple statistics, 1st, 2nd, 3rd and 4th order moments
- growing of trainingsets for classification
Johan Doré Hansen)
- semivarmodel - estimate semivariogram model based on experimental semivariogram
- sep2dfilter - calculate whether a 2-D filter is separable and if so, separate
- specInfoDiv - calculate spectral information divergence (i.e., the symmetrized Kullback-Leibler divergence or relative entropy) between all spectra in two image cubes
- specInfoDivRef - calculate spectral information divergence (i.e., the symmetrized Kullback-Leibler divergence or relative entropy) between all spectra in an image cube and a reference spectrum
- sigma_n - estimate noise covariance matrix of irregularly spaced data
- specInfoMeas - calculate spectral information measure (i.e., the entropy) of all spectra in an image cube
- spectral angle mapper
- spectral angle change detection
- standardize float image to desired mean and stddev
- target constrained interference minimization filter (TCIMF)
- full and partial spectral unmixing
- Wallis filter
- 1-D, 2-D or 3-D wavelet transformation
- test for equality in two comlex polarimetric radar signals
(can be used to carry out change detection in polarimetric SAR data)
- wishart_det - calculate determinants of all pixels in covariance representation of polarimetric SAR image data
- wishart_change - based on the complex Wishart distribution calculate change between two polarimetric SAR images in the covariance representation
- wishart_change_dualk - based on the complex Wishart distribution calculate change between several dual pol SAR images in the covariance representation
that relate to this type of
(exploratory) data analysis
written by colleagues and students
plus several others (which may be added later).
- exploratory projection pursuit
- grand tour
- non-linear canonical correlations analysis via ACE
- gray level cooccurrence matrices
(J. Michael Carstensen)
- lintrans - linear/affine transformations of multivariate data
- mace - non-linear multiset canonical correlations analysis via ACE
(Klaus Baggesen Hilger)
(The information below is a little backdated).
J. Michael Carstensen
I maintain a collection of IMM written programs
for analysis of spatial and image data.
The programs come in two groups, one of which is freely distributed.
All programs run under UNIX and comply with the
HIPS comes with source code and is very open
and easily extended with your own software.
List of freely distributed HIPS programs
from the IMM Section for Image Analysis.
(These programs are good with HIPS only and they are distributed with HIPS
at the time of purchase.)
List of other HIPS programs
from the IMM Section for Image Analysis.
See my homepage.