Page 174 - Digital Analysis of Remotely Sensed Imagery

Image Pr ocessing Systems 139

all systems offer nearly the same suite of functions as their competitors.
Thus, they are losing their unique individuality. In terms of workflows,
the processed results are integrated into other systems or, increasingly,
with other GIS layers for further analysis in the same image analysis
system.
These systems are compared with each other in several categories
under image display, data preparation and image enhancement,
image classification, and user interface in Table 4.1. eCognition and
GRASS are excluded from the comparison because they are not
generic image analysis systems. Originated from the personal com-
puter background, IDRISI is best at teaching digital image process-
ing, owing to its modular structure and its comprehensive range of
image classifiers and change detection analysis tools. Primarily, it is
still a digital image analysis system with GIS functions. This sys-
tem has a limited capacity to accurately georeference images using
sensor-specific models. It also lacks the capability of analyzing stereo
images and producing DEMs. Processing speed is slow for large
image files. It has reached the industry standard, but not for all types
of image analysis.
In comparison with other image processing packages, ERDAS
Imagine is the most capable of handling vector data and integrating
image analysis with GIS (e.g., ESRI ArcGIS). After being acquired by
Leica, ERDAS has expanded its functionality in analyzing stereo-
scopic images and extracting topographic information from remote
sensing materials. Its range of image classification methods has broad-
ened with the addition of the subpixel classifier and modules for clas-
sifying hyperspectral data. However, in the recent release it still lacks
image classification capability based on machine learning or pixel
spatial properties. Despite being the only system that is able to under-
take intelligent image analysis.
ENVI used to be the only system able to process hyperspectral
data, but this advantage is disappearing quickly, as other systems
also offer similar modules. Now its functionality has expanded to
such a degree that its gap with ERDAS is being bridged up very
quickly. Consequently, it shares a number of similarities with ERDAS
Imagine. For instance, both are designed with the purpose of per-
forming all kinds of image analysis comprehensively. ENVI is a slight
leader in handling hyperspectral data. Its other strength is the ability
to offer more image classifiers than ERDAS (but less than IDRISI), as
well as a broad range of texture descriptors in spatial image analysis.
With the release of the Feature Extraction add-on module, ENVI is
the only mainstream system able to perform object-oriented image
classification. However, it is disadvantaged by its limited capacity in
processing 3D data. The only non-remote sensing data ENVI can han-
dle are topographic in nature. ENVI is way behind ERDAS in pro-
cessing vector data, and in integrating image analysis with GIS and
photogrammetry.