SpectralView is a PC-based application for visualizing and analyzing multi- and hyperspectral images. It is engineered to read data cubes consisting of raw pixel counts, and perform necessary conversion to engineering units. SpectralView provides a number of simple tools for plotting and representing data, and complex and powerful tools for analyzing principal components and data clustering.
view full-size screenshot
The visualization and analysis functions described in this document are
available in the standard version of SpectralView. The software is also
available customized to a specific application based on data cube format,
source or type of target, and background information. Contact Signature
Research for more information about customization options.
The spectral paging feature allows the user to either
page through the various spectra individually with a
click of a button, or view an animation of the entire
hyperspectral data set.
Spectral Angle Matching
SpectralView displays binary or gray scale threshold
images based on the L2 inner product between a
target spectra and a hypercube.
The data clustering tool allows the user to view the
inherent spectral anomalies between scene elements.
It automatically partitions the data into clusters
whose spectra are fundamentally similar.
SpectralView allows the user to either choose from one
of three methods of initialization, or let SpectralView
find a clustering configuration based on a principal
components analysis. Once a set of clusters is
generated, the user can view the clusters either in the
original image or as a two-dimensional scatter plot
corresponding to the first two principal axes.
SpectralView can generate RGB images and perform
spectrum-to-CIE coordinate and xyY-to-RGB
conversion from spectra in the input data. CIE color
coordinates for the pixel under the cursor are plotted
in real-time, including a marker for the spectral
position of the view. SpectralView also allows the user
to view color differences based on a user-defined AOI.
Principal Components Analysis
SpectralView can find the projection of the hypercube
data onto the principal axes using spectral
decomposition. The user can then view the resulting
images, which are linear combinations of the original
hypercube images, with the first representing the
image with the maximal statistical variance. This
tool acts as a data compression technique and allows
the user to perceive the geometric distribution of the
original multi-dimensional data.
In many research applications, it is important to define
and analyze discrete regions of the image - Areas of
Interest (AOIs) - very precisely. SpectralView has
unique AOI drawing, editing, and storage functions.
The editing function allows the user to graphically add
or delete nodes and to move the nodes independently
on an enlarged view of the image. A group move
function simultaneously aligns all of the AOIs while
maintaining the relative positions. The user can also
name and store each AOI for later use.
The data export feature allows the user to export to
an MS Excel-compatible file the colorimetric
information for AOIs within the RGB images
generated. Individual image data occupies one line in
the file to facilitate processing multiple images.