Page 159 - Digital Analysis of Remotely Sensed Imagery

126 Cha pte r F o u r

elevation on the display is shown immediately and interactively by
moving the shading angle or enhancement saturation button. It is
possible to display multiple layers of vector data from different
sources in different formats on top of raster imagery. The displayed
image may be manipulated using the transform editor. Processed
results can be viewed instantly using the display and mosaic wizard
that decides the best setting for the display after it detects the type of
the data. An image displayed as 2D views may be changed to 3D
perspective or 3D fly-through views if the DEM of the area is avail-
able. Viewing angle and zoom factor in 3D perspective can be interac-
tively modified.
4.4.3 Image Enhancement and Classification
In ER Mapper the contrast of images may be adjusted linearly or
piecewise linearly, using gaussian and histogram equalization. They
can also be manipulated using a full range of transforms, such as
principal component analysis, Tasseled Cap transformation, RGB to
HIS transformation, and fast Fourier transformation. Multiple bands
may be ratioed. Images may also be spatially enhanced using convo-
lution filtering of high pass, low pass, edge enhancement, adaptive
median, morphological, and majority filtering.
ER Mapper is able to perform both unsupervised and supervised
classifications. The analyst has full control over the unsupervised classi-
fication by specifying certain parameters, such as maximum standard
deviation within a class, separation value when splitting classes, and
minimum distance between class means. All common supervised classi-
fiers, such as parallelepiped, minimum distance, maximum likelihood,
and mahalanobis, are supported in ER Mapper. Unlike its counterpart in
other systems, the maximum likelihood classifier in ER Mapper incorpo-
rates contextual information gleaned from neighboring pixels into its
decision making. Knowledge on the identity of the surrounding pixels is
used to help smooth the classification. The classified results are evalu-
ated against the reference data or other classified images for their accu-
racy. The produced confusion matrix shows the producer’s and user’s
accuracy, as well as the overall accuracy. The classified raster imagery
may be converted automatically into a polygon coverage using the pow-
erful vectorization tool.
4.4.4 Image Web Server
Perhaps the most distinct module of ER Mapper Enterprise is its image
web server. With the increasing popularity of the Internet, images are
delivered to remote users electronically. How to share images among
different researchers working in a larger project via the Internet
and intranet is an issue that caught the attention of ER Mapper earlier
on. Their response to this demand is the image web server that is
designed for fast delivery of large raster image files over the Internet
and intranets. One version of the server bundled with ER Mapper 7.0