50
J.Jensen
retain sufficient detail for orientation purposes. For a 9 x 9" photograph, this produces a 0.81 megabyte file (9 x 9 x 100 x 100 dpi = 810,000 bytes) for black and
white photographs and a 2.43 MB file for color aerial photography. This assumes
that the black and white images are scanned at 8-bit radiometric resolution and
that color images are scanned at 24-bit resolution. Images that are to be used for
scientific purposes are usually scanned at much higher spatial resolution such as
500 dpi (12.7 ~m) to retain the subtle reflectance and/or emittance information
found in the original imagery. For 1 :20,000 scale photography, this yields a spatial
resolution of 1.02 x 1.02 m per pixel and results in a 20.25 MB file for a single
black and white photograph and a 60.75 MB file for a single color photograph
(Jensen, 1995b; 1996).
Most aerial photography is collected using a 9 x 9" metric camera. The ideal
scanning system digitizes the entire image and any ancillary 'titling' information
on the periphery of the image at one time. Therefore, the digitizer of choice should
have a field of view of at least 9 x 9". While it is possible to use inexpensive 8.5 x
11" desktop scanners, this requires that the imagery be broken up into two parts
that must then be mosaicked together. This introduces radiometric and geometric
error and should be avoided. Advances in the 'desktop publishing' industry have
spurred the development of flatbed, 11 x 14" desktop linear array digitizers that
can be used to digitize negatives, diapositives, and paper prints at 100 to 3,000
dots-per-inch (Foley et aI., 1994). Some scanners digitize color photographs with
one color filter then repeat the process with the other color filters. This can result
in color misregistration and loss of image quality. Ideally, color aerial photography is scanned in a single pass and converted into three registered red, green, and
blue (RGB) files.
Area array charge-coupled-device (CCD) digital camera technology has been
adapted for hard-copy image digitization (Fig. 3.1). Typical area array CCD systems digitize from 160 dpi to 3,000 dpi (approximately 160 ~m to 8.5 ~m) over a
10 x 20" image area (254 ~m x 508 ~). They scan the original negative or positive transparency as a series of rectangular image tiles. The scanner then illuminates and digitizes a reseau grid which is an array of precisely located crosshatches etched into the glass of the film carrier. The reseau grid coordinate data
are used to locate the exact orientation of the CCD camera during scanning and to
geometrically correct each digitized 'tile' of the image relative to all others. Radiometric calibration algorithms are then used to compensate for uneven illumination encountered in any of the tile regions. Area array digitizing technology has
obtained geometric accuracy of < 5 ~ over 23 x 23 cm images when scanned at
25 ~ per pixel and repeatability of < 3 ~ (Jensen, 1996).
3.2 Image Processing and GIS Software Requirements
A variety of digital image processing and geographic information system (GIS)
functions are required to analyze remotely sensed data for hydrology and water
resource management applications. Some of the most important functions are
summarized in Table 3.2. It is useful to briefly identify characteristics of the most
J.Jensen
retain sufficient detail for orientation purposes. For a 9 x 9" photograph, this produces a 0.81 megabyte file (9 x 9 x 100 x 100 dpi = 810,000 bytes) for black and
white photographs and a 2.43 MB file for color aerial photography. This assumes
that the black and white images are scanned at 8-bit radiometric resolution and
that color images are scanned at 24-bit resolution. Images that are to be used for
scientific purposes are usually scanned at much higher spatial resolution such as
500 dpi (12.7 ~m) to retain the subtle reflectance and/or emittance information
found in the original imagery. For 1 :20,000 scale photography, this yields a spatial
resolution of 1.02 x 1.02 m per pixel and results in a 20.25 MB file for a single
black and white photograph and a 60.75 MB file for a single color photograph
(Jensen, 1995b; 1996).
Most aerial photography is collected using a 9 x 9" metric camera. The ideal
scanning system digitizes the entire image and any ancillary 'titling' information
on the periphery of the image at one time. Therefore, the digitizer of choice should
have a field of view of at least 9 x 9". While it is possible to use inexpensive 8.5 x
11" desktop scanners, this requires that the imagery be broken up into two parts
that must then be mosaicked together. This introduces radiometric and geometric
error and should be avoided. Advances in the 'desktop publishing' industry have
spurred the development of flatbed, 11 x 14" desktop linear array digitizers that
can be used to digitize negatives, diapositives, and paper prints at 100 to 3,000
dots-per-inch (Foley et aI., 1994). Some scanners digitize color photographs with
one color filter then repeat the process with the other color filters. This can result
in color misregistration and loss of image quality. Ideally, color aerial photography is scanned in a single pass and converted into three registered red, green, and
blue (RGB) files.
Area array charge-coupled-device (CCD) digital camera technology has been
adapted for hard-copy image digitization (Fig. 3.1). Typical area array CCD systems digitize from 160 dpi to 3,000 dpi (approximately 160 ~m to 8.5 ~m) over a
10 x 20" image area (254 ~m x 508 ~). They scan the original negative or positive transparency as a series of rectangular image tiles. The scanner then illuminates and digitizes a reseau grid which is an array of precisely located crosshatches etched into the glass of the film carrier. The reseau grid coordinate data
are used to locate the exact orientation of the CCD camera during scanning and to
geometrically correct each digitized 'tile' of the image relative to all others. Radiometric calibration algorithms are then used to compensate for uneven illumination encountered in any of the tile regions. Area array digitizing technology has
obtained geometric accuracy of < 5 ~ over 23 x 23 cm images when scanned at
25 ~ per pixel and repeatability of < 3 ~ (Jensen, 1996).
3.2 Image Processing and GIS Software Requirements
A variety of digital image processing and geographic information system (GIS)
functions are required to analyze remotely sensed data for hydrology and water
resource management applications. Some of the most important functions are
summarized in Table 3.2. It is useful to briefly identify characteristics of the most