3 Processing Remotely Sensed Data: Hardware and Software Considerations
49
spective when perfonning terrain analysis using remote sensor data. Therefore, the
image processing system should be able to display at least 512 rows x 512 columns of pixels and preferably more (e.g., 1024 x 1024) on the CRT at one time.
This allows larger geographic areas to be examined at one time and places the
terrain of interest in its regional context.
3.1.7 Screen Color Resolution
CRT screen color resolution is the number of gray-scale tones or colors (e.g., 256)
that a pixel may be displayed in on a screen out of a palette of available colors.
Most sophisticated image processing systems provide a tremendous number of
displayable colors from a large color palette (e.g., 16.7 million). The primary
reason for these significant color requirements is that image analysts must often
display a color composite of several individual images at one time on the screen.
For example, to display the false-color composite of Hurricane Andrew in the
Gulf of Mexico (Colour Plate 3.A) it was necessary to place three separate 8-bit
images in three distinct planes of image processor memory [A VHRR band 5 (11.5
- 12.5 Ilm) thennal infrared data were placed in the blue image processor memory
plane; band 2 (0.725 - 1.10 J..lffi) near-infrared data were placed in the green memory plane; and visible band 1 (0.58 - 0.68 J..lffi) data were placed in the red image
plane]. Thus, each pixel could take upon itself any of 224 possible color combinations (16,777,216). Such true color systems are expensive because every pixel
location is bit mapped, i.e. a specific location in memory keeps track of the blue,
green, and red brightness values.
The network configured in Fig. 3.1 has six 8-bit color workstations and four 24bit color workstations. Everyone does not require access to a 24-bit color display
at once because many image processing functions such as black and white image
display and the creation of final color thematic maps can be perfonned quite well
in 8 bits (Busbey et aI., 1992).
3.1.8 Image Scanning (Digitization) Considerations
Many hydrologic projects require the analysis of analog (hard-copy) aerial photography such as color-infrared U.S. National Aerial Photography Program (NAPP)
imagery. Also, many studies make use of multiple dates of historical panchromatic
black and white aerial photography (e.g. 1 :20,000 U. S. Agricultural Stabilization
and Conservation Service photography obtained in the 1930s and 1940s). Such
data are often digitized, rectified, and then analyzed in a digital image processing
system. To be of value, careful decisions must be made about the spatial and radiometric resolution of the digitized image data. Tables that summarize the relationship between input image scale, digitizer detector instantaneous-field-of-view
(IFOV) measured in dots-per-inch (dpi) and micrometers (J..lffi), and output image
spatial resolution in meters are found in Jensen (1995b; 1996).
Images to be used for simple illustrative purposes may be scanned at low resolutions such as 100 dpi (254 J..lffi). For vertical aerial photography obtained at
1:20,000 scale, this yields images with a spatial resolution of 5.08 x 5.08 m that
49
spective when perfonning terrain analysis using remote sensor data. Therefore, the
image processing system should be able to display at least 512 rows x 512 columns of pixels and preferably more (e.g., 1024 x 1024) on the CRT at one time.
This allows larger geographic areas to be examined at one time and places the
terrain of interest in its regional context.
3.1.7 Screen Color Resolution
CRT screen color resolution is the number of gray-scale tones or colors (e.g., 256)
that a pixel may be displayed in on a screen out of a palette of available colors.
Most sophisticated image processing systems provide a tremendous number of
displayable colors from a large color palette (e.g., 16.7 million). The primary
reason for these significant color requirements is that image analysts must often
display a color composite of several individual images at one time on the screen.
For example, to display the false-color composite of Hurricane Andrew in the
Gulf of Mexico (Colour Plate 3.A) it was necessary to place three separate 8-bit
images in three distinct planes of image processor memory [A VHRR band 5 (11.5
- 12.5 Ilm) thennal infrared data were placed in the blue image processor memory
plane; band 2 (0.725 - 1.10 J..lffi) near-infrared data were placed in the green memory plane; and visible band 1 (0.58 - 0.68 J..lffi) data were placed in the red image
plane]. Thus, each pixel could take upon itself any of 224 possible color combinations (16,777,216). Such true color systems are expensive because every pixel
location is bit mapped, i.e. a specific location in memory keeps track of the blue,
green, and red brightness values.
The network configured in Fig. 3.1 has six 8-bit color workstations and four 24bit color workstations. Everyone does not require access to a 24-bit color display
at once because many image processing functions such as black and white image
display and the creation of final color thematic maps can be perfonned quite well
in 8 bits (Busbey et aI., 1992).
3.1.8 Image Scanning (Digitization) Considerations
Many hydrologic projects require the analysis of analog (hard-copy) aerial photography such as color-infrared U.S. National Aerial Photography Program (NAPP)
imagery. Also, many studies make use of multiple dates of historical panchromatic
black and white aerial photography (e.g. 1 :20,000 U. S. Agricultural Stabilization
and Conservation Service photography obtained in the 1930s and 1940s). Such
data are often digitized, rectified, and then analyzed in a digital image processing
system. To be of value, careful decisions must be made about the spatial and radiometric resolution of the digitized image data. Tables that summarize the relationship between input image scale, digitizer detector instantaneous-field-of-view
(IFOV) measured in dots-per-inch (dpi) and micrometers (J..lffi), and output image
spatial resolution in meters are found in Jensen (1995b; 1996).
Images to be used for simple illustrative purposes may be scanned at low resolutions such as 100 dpi (254 J..lffi). For vertical aerial photography obtained at
1:20,000 scale, this yields images with a spatial resolution of 5.08 x 5.08 m that