IDL Extensions for ENVI

Mort Canty

This page lists the most recent versions of my IDL programs for the ENVI environment discussed in my textbook Image Analysis, Classification and Change Detection in Remote Sensing, with Algorithms for ENVI/IDL,Second Revised EditionTaylor & Francis, CRC Press 2010as well as some additional, mostly experimental, routines for parallel processing with Nvidia'sCUDA and Python versions of some of the algorithms.

This page was last modified 05/16/2013 13:01:43.
- Included an ENVI/IDL extension for Scatterplot normalization

Return to my private homepagehere.

If you don't have ENVI/IDL, you can try out some of the algorithms on theGoogle Cloud.See also Allan Nielsen's software page for Matlab versions of the change detection algorithmsand my Python versions of IR-MAD, Radiometric Normalizationand kernel PCA.

Prerequisite Libraries

The following libraries must be present in the IDL path before attempting to run any of the extensions:

David Fanning's Coyote Library

My auxilliary routines (Documentation.)

On 32bit Windows systems place PROVMEANS.x86.DLL and PROVMEANS.DLM from this library in your DLM path.

Note: If you are running IDL on Windows x64, place PROVMEANS.x86_64.DLL and PROVMEANS.DLM in the DLM path.In addition, if you don't have Microsoft Visual C installed on your computer,you will need to download and install the Microsoft Visual C 2010 Redistributable Package (x64).

If you're not running on Windows, see the textbook for instructions.

All extensions also assume that ENVI is up and running.Most of them can be integrated directly into the ENVI main menu by copying the programswith filenames of the form program_RUN.PRO to ENVI's SAVE_ADD directory.

In addition some of the extensions can take advantage of the Tech-X Corp.GPULib interface to CUDA.(These extensions will now also run without GPULib/CUDA.)

Documentation

Kernel K-Means Clustering of Remote Sensing Imagery with CUDA, a detailed description of my GPULib implementation of the Kernel K-Means algorithm (Google Docs).

The MAD MAN, a users manual for the IR-MAD (iMAD) and RADCAL extensions (Google Docs).

ENVI/IDL Extensions

Python routines

Downloads

Note: If you use this software, please do not forget to acknowledge the source. If you use the extension(s) for IR-MAD (iMAD) analysisplease cite Allan Nielsen's IR-MAD paper,if you use the extension(s) for IR-MAD radiometric normalization please cite thenormalization paperby myself and Allan Nielsen, if you use the kMAF extension please cite Allan's kMAF/kMNF paper.

Preprocessing	DWT fusion	sharpen multispectral images with discrete wavelet transform
	A trous fusion	ditto with a trous wavelet transform
	Wang-Bovik quality index	evaluate radiometric fidelity of pansharpened images
	C-correction	correct for solar illumination in rough terrain
	Kernel PCA	perform nonlinear principal components analysis (can take advantage of GPULib)
	Kernel MAF	perform nonlinear maximum autocorrelation factor analysis (can take advantage of GPULib)
	Contour-match	get tie-points for image-image registration from invariant features
Supervised classification	Bayes maximum likelihood	wrapper for the ENVI ML classifier
	Support vector machine	wrapper for the ENVI SVM classifier
	Hybrid two-layer neural network	trained with kalman filter and scaled conjugate gradient algorithms
	Two-layer neural network	trained with scaled conjugate gradient algorithm (can take advantage of GPULib)
	Boosted three-layer neural network	apply adaptive boosting (AdaBoost) to a sequence of neural networks
	Gaussian kernel classification	non-parametric Parzen-window classification (can take advantage of GPULib)
	Probabilistic label relaxation	perform postclassification filtering
	Contingency table	calculate confusion matrices and kappa values
	McNemar test	compare classifiers with the McNemar statistic
Unsupervised classification	Expectation maximization	cluster image data with a mixture of multivariate Gaussians (can take advantage of GPULib)
	FKM clustering	cluster image data with a fuzzy K-means algorithm
	HCL clustering	cluster image data with a heirarchic agglomerative algorithm
	Kernel K-means	cluster image data with a kernel version of K-means (can take advantage of GPULib)
	Kohonen SOM	visualize image data with the Kohonen self-organizing map
	Mean shift	segment images with mean-shift algorithm
Change detection	IR-MAD (iMAD)	apply iteratively re-weighted multivariate alteration detection
	Radcal	perform automatic relative radiometric normalization of images
	MadView	set thresholds on MAD images
	Bslfcpnorm	Scatterplot normalization of RED and NIR image bands
Miscellaneous	Structure height	use RFMs to determine height of vertical structures
	Class segmentation	Segment a classified image
	Examples	example IDL programs from the 2nd edition
	Solutions	some solutions to the progamming exercises in the 2nd edition
CUDA (experimental)	Cuda_SVD	a DLM for singular value decomposition on CUDA
	Cuda_NDVI	a DLM for calculating NDVI indices on CUDA
	Cuda_STRETCH	a DLM for enhacement stretching on CUDA
Python	iMad/RADCAL	Python scripts for IR-MAD (iMAD) and RADCAL
	kPCA	Python script for kernel PCA with and without CUDA