Page 135 - Digital Analysis of Remotely Sensed Imagery

106 Cha pte r F o u r

processing, this overview thus concentrates mostly on the image anal-
ysis components of these systems. The strengths and limitations of
each system in image analysis will be critically evaluated and com-
pared with other systems wherever possible.

4.1 IDRISI
IDRISI is a sophisticated desktop raster geographic information and
image processing system developed by the Graduate School of
Geography at Clark University, Worcester, Massachusetts. The lat-
est release, called IDRISI Andes (version 15), is 32-bit Windows NT–
compatible. This affordable system comprises over 250 modules or
stand-alone programs for the digital analysis and visualization of
spatial data, including remotely sensed imagery, in a single pack-
age. These modules ranging from the basic to highly advanced in
their functionality, and are grouped into database query, spatial
modeling, image enhancement, and classification. Those modules
related specifically to GIS, such as database query and GIS model-
ing, will not be covered here. Instead, this section focuses on its
image analysis functions.

4.1.1 Image Analysis Functions
The capacity of IDRISI for processing remotely sensed data falls into
six areas: image restoration, enhancement, transformation, classifi-
cation, change detection, and accuracy assessment. In image resto-
ration, images are corrected both geometrically and radiometrically
using the procedures in IDRISI. Radiometric correction may be
undertaken to eliminate the atmospheric effects and destripping.
Images can be geometrically corrected using interactively selected
ground control points (GCPs). Such images may be integrated with
georeferenced data from other sources. Images may be enhanced
via contrast adjustment, PAN sharpening (i.e., merging of the pan-
chromatic band with the multispectral bands from the same sensor),
and filtered using edge enhancement. The spectral quality of an
image can be enhanced using such modules as noise removal
through convolutional filters and Fourier analysis. IDRISI provides
all major data preparatory tools, such as image subsetting, mosaick-
ing and vector generalization. Images may be transformed using an
extensive range of procedures that include principal component
analysis, canonical component analysis, color space transforma-
tions, and vegetation indexing.
IDRISI offers an unparalleled suite of classifiers among all lead-
ing image analysis systems. Remote sensing data can be classified
either unsupervised or supervised. The unsupervised method is
based on clustering analysis. The supervised classifiers include maxi-
mum likelihood, minimum distance to means, and parallelepiped.