2.0 INTRODUCTION
Many of the image pre-processing steps are conducted within an image-processing software package. The Center for Coastal Geology group uses PCI, Inc. The following chapter includes specific instructions on running the procedures. It should serve as a guide to the basic steps required within this program.
Initial scene selection and ordering is
conducted outside of the image processing
package. A flow chart
(Plate 4)
is provided to
help the novice follow the different kinds of steps
involved and the order in which they are
conducted. The check list in Appendix A
should become the basis for your own notes,
including adjustments and new developments.
2.1 SCENE PREPARATION
Imagery is downloaded from the CD-ROM
disk with CDNLAPS or CDEOSAT,
depending on the format of the disk. Important
information is included in the header file. Take
time to learn to read the header file with the
instructions accompanying your order. Look for
format (NLAPS or EOSAT), band sequential or
pixel interleaved (BSQ, BIL), number of pixels
and lines, date and time of acquisition,
resampling scheme (NN, CC, BL), and so on.
Once downloaded to a .pix file, examine each
band at full resolution for missing lines or blocks
of missing data, heavy clouds, and the histogram
of dn values. Each band should exhibit a normal
range and distribution for the season and region
covered. Now is the time to repair or return an
image with serious problems.
Table 3 shows the parameters set for
CDNLAPS. Table 4 shows parameters set for
LRP to fix line number 100 in band 2 with the
mean from the line above and the line below.
The EASI procedure, LRP, requires the operator
to specify the missing line(s) +1 to correctly
identify the lines to be replaced. If destriping is
attempted, read about DSTRIPE in EASI help.
2.2 ORTHORECTIFICATION OF IMAGERY
The orbital segment created in
CDEOSAT or CDNLAPS contains information
used in an adjustment between the estimated
latitude/longitude of the raw imagery and the
desired coordinate system. SMODEL creates a
model from the orbital segment and a ground
control point segment. The ground control
segment is created in GCPWORKS.
2.2.1 GCPWORKS
Imagery for the Florida Wetlands is
rectified with ground control collected with GPS
units and is reprojected to the UTM coordinate
system. Ground control selection is described in
Chapter 1, and can be conducted in any image
display program, including PCI's
IMAGEWORKS. Once the ground control has
been identified and field collected, run
GCPWORKS to enter the coordinates at their
respective locations in the image. The satellite
orthorectification option is selected to permit the
use of the orbital segment in the model.
Select "Collect/review GCPs only",
"Satellite Ortho Correction", "User Entered
Coordinates", and then "Select Uncorrected
Image". Enter the path and filename of the
nlaps.pix file. Load three bands, usually natural
color (3,2,1) or false color (4,3,2) are preferred
for display. Set the units to meter, and the
coordinate system to UTM, zone 17, row R and
ellipsoid WGS84. Obviously, the zone and row
will change depending on the region. Identify
the location of 8 to 12 ground control points and
enter the corresponding UTM coordinates
collected with the GPS unit.
During the identification of ground
control within GCPWORKS, watch the RMS
error for each ground control point and the total
RMS in both x and y. Within reason, it is
possible to adjust the location of the ground
control in the imagery to achieve optimal
positioning and distortion-free rectification.
In GCPWORKS it is possible to also
enter accuracy check points at this time. If not
working in this program, these check points are
set aside for a post-rectification accuracy check.
The 10-12 accuracy check points must be of
equally high quality and well-distributed about
the scene. Care is given to treat these positions
and the associated error objectively, as they
eventually provide a measure of the model's
accuracy. These positions should be entered at
the end, when the ground control points have
been securely identified to avoid adjusting the
model to fit the check points. When the operator
is satisfied with the location of the ground
control and the total RMS error, the ground
control point segment is saved within
GCPWORKS. Ideally, this will be segment #3.
2.2.2 GCPREP
A report documenting the ground
control point segment as collected is obtained by
running GCPREP. This produces a text file
showing the list of ground control points, their
pixel/line locations, the user-entered coordinates,
and errors. The UTM coordinates of the ground
control is exactly what was entered within
GCPWORKS. This will change when the
model is run in SMODEL. Remember, in
GCPWORKS, the satellite orthorectification was
selected. This forces the orbital segment to be
referenced while developing the model.
GCPREP does not refer to the orbital segment,
and thus does not reflect it's influence on the
ground control. The output from another
procedure, GCPWRIT, may be useful in creating
a vector segment. It requires a little editing and
then running VREAD. The vector may be
displayed in subsequent displays to help relocate
the ground control points for new imagery.
2.2.3. SMODEL
SMODEL draws on both the orbital
segment and the ground control point segment to
create the model, and must always be run on the
whole scene. It is ineffective to run on a partial scene.
If check points were entered
along with the ground control, SMODEL will
also produce a table showing the accuracy of the
check points in meters. Run SMODEL as
shown in Table 5. Both the orbital segment, 2,
and the ground control point segment, 3, must
be specified in the procedure. SMODEL
produces a model segment, normally 4, which
will be applied in the image rectification,
SORTHO.
SMODEL also produces a report
showing the accuracy of the check points in
meters. Examine the report for values outside of
the range of expected and desired accuracy (see
Appendix C). Excessive values below the
original 10 m accuracy of the GPS positions may
introduce distortion to the final image
rectification. Values greater than 20 m, or 2/3
pixel, suggest the overall registration may not
meet inter-scene registration requirements. Note
also that the UTM coordinates shift by the RMS
error.
Values for positional accuracy in this
project are acceptable between 8-20 m. The
rectification of a TM scene will technically meet
1:25,000 map accuracy standards, provided that
90% or more of the individual accuracy check
points are within 20 m or less of their desired
location. With 12 check points, only one can be
2.2.4 CIM2 - New file creation
New imagery files are created in CIM2,
a file format which creates a separate file for each
band, all associated by the header or .pix file
(Table 6). The size of a fully-processed full size
Landsat TM scene makes this option appealing
for ease of movement and disk-space issues. The
example shows the standard size of the north Big
Bend image, path 18, row 39. The size includes
the removal of ~1100 lines in the state of Georgia
and the enlargement of the file to reproject the
region into the UTM coordinate system (see
Figure 3). All north scenes will be created in the
same manner. If working in another region,
establish a standard file size based on the full
extent of the UTM eastings and northings and
the pixel size.
2.2.5 GEOSET - georeference segment
Every file in PCI has a georeferencing
segment created automatically as segment 1.
The geographic boundaries, the UTM zone, and
the ellipsoid are set for the georeferencing
segment of the newly created CIM2 file
with GEOSET. An example for the
north Big Bend is shown in Table 7. The
coordinates given are applied to each
corresponding image, making overlays and inter-
scene analysis a simple task.
2.2.6 SORTHO - satellite image
orthorectification
Orthorectification of imagery is
conducted in PCI with the SORTHO procedure
which applies the model segment to the nlaps
image file and reprojects the seven bands of data
into the empty bands of the newly created and
georeferenced .pix file. See Table 8 for
SORTHO parameters.
Due to the minimal topographic
variation along the Big Bend coast of Florida,
we do not employ a DEM file. Instead, we
estimate an average value for elevation as the
width of one pixel with no offset (ESCALE).
Elevations do not exceed ~60 m in the area, and
this appears to be sufficient for positional
accuracy and inter-scene registration. The need
for elevation must be determined on a region-by-
region basis. The SORTHO procedure may
take up to 6 hours on a SUN SPARC 20 and is
usually run overnight.
2.2.7 Inter-scene accuracy check
The inter-scene registration accuracy
check is conducted manually within an image
display program. Run IMAGEWORKS with a
full-resolution window ~600 x 800 with 6 image
planes. Display 3-5 well-distributed regions in
the window, each time loading the newly
registered image (4,3,2) and the base image
(4,3,2) with the same input window.
For instance, in image planes 1,2,3 load
the new image (4,3,2) with an input window of
300, 300 (x,y offsets), 600, 800 (x,y window
size). Then load the base image to image planes
4,5,6 (4,3,2) using the same full-resolution
window. Examine the images first with a single
band from each of the images in a RGB display
mode. An example may be viewed in
Plate 3.
Look for truly horizontal or vertical roads and
evaluate obvious offsets. A poorly registered
image as shown to the left in
Plate 3 is not
worth evaluating further. Note the direction of
the offsets, check other regions to confirm the
offsets, and return to GCPWORKS to redo the
ground control point segment.
If horizontal and vertical roads align
well in the first input window, switch to a flicker
state with each image displayed 4,3,2 as RGB.
Select 3-5 right-angle road intersections for
evaluation. See
Plate 2
for selection of road
intersections. Flicker to the base image and
zoom in on a selected intersection. Place the
cursor at the center of the intersection, drawing
two imaginary lines N/S and E/W to determine
the exact location. Leaving the cursor in
position, flicker to the newly registered image.
Again, mentally draw two lines and determine
the exact center of the new intersection. Do not
move the cursor. Evaluate if the cursor is exactly
on target, off by 0.5 pixel, 1.0 pixel, or > 1.0
pixel in both the x and y directions. See
Plate 1
for evaluation of cursor location.
Complete a table as follows, with 12-20
road intersections across the scene. The table
will highlight regional or full-scene offsets. If all
regions, and the whole scene meet expectations
of ± one pixel inter-scene alignment, then the
new file is acceptable, and image processing
proceeds. If, however, a region or the whole
scene exceeds one pixel offset in either direction,
then GCPWORKS, SMODEL and SORTHO
must be repeated. The table may be used to help
guide the adjustment of the ground control
points.
2.3 RADIOMETRIC AND
ATMOSPHERIC CORRECTION
The following sections illustrate the
EASI procedures used for radiometric calibration
and atmospheric correction. Information from the
header file is employed in the procedures. A
bitmap is created first to mask the non-image
pixels of the *ort.pix file. Typically the mask
can be obtained by selecting all pixels in band 1
or band 2 with values greater than zero. The mask
is subsequently used for the creation of indices
and other analysis. Limiting the processing to
the image area of the file shortens processing
time and eliminates the formation of anomalous
features outside of the image area. The EASI
thresholding procedure, THR may be used for
creating the bitmap (Table 9). If necessary a
similar mask may be created for the non-image
area by setting the COMP parameter to "ON".
The complement is used to fix anomalous data
sometimes found in band 6.
2.3.1 TMRAD
Seven additional empty channels are
added using PCIADD2. These are the output
channels for the radiometrically enhanced bands,
8-14. TMRAD is an in-house EASI procedure
which solves equations 1 and 2 in section 1.4.1.
Sample parameters are set in Table 10.
Latitude and longitude are given in decimal degrees, and
longitude is a negative value in the western hemisphere.
TMSLOPE and TMBIAS may be obtained from the newer header
files as Gain/10 and Bias/10, respectively. The solar constants
(from Markham and Barker, 1986) and scale factors will
change for different imagery and satellites. TEXT may
be set to "stat" for a check on the output values
Table 11, which is also
saved to a text file for reference.
2.3.2 RAYRAD - Rayleigh radiance
The Rayleigh correction factor is
calculated with RAYRAD. Calculations may be
found in Stumpf (1992). The procedure requires
operator input as in Table 12, and produces
output as in Table 13. The scale factor should
be the same as that used in TMRAD.
Bands 1 - 4 are corrected by subtracting
the Rayleigh values from each band respectively.
The Rayleigh values are the last column in Table 13,
rounded to the nearest whole number. Values for
bands 5 and 7 are insignificant, and for band 6,
irrelevant.
The model used to accomplish the subtraction is
shown in Table 14.
The bitmap is specified with "%%2" and band 8
as "%8".
Keep in mind that the
radiometrically enhanced bands, 8 - 14, are
equivalent to TM bands 1 - 7.
2.3.3 Dark object subtraction
Table 3. CDNLAPS procedure parameters
CDEOSAT CD EOSAT Fast Format V6.0 EASI/PACE 12:31 16-Feb-97
NLAPSHD - NLAPS Header File Name :/cdrom/cdrom1/filename.hd
FILE - Database File Name :path/newfile.pix
CDIC - CD Input Channel List } 1 2 3 4 5 6 7
TEX1 - Database Descriptive Text 1 :path_row_location
REPORT - Report Mode: TERM/OFF/filename :path/newfileorb.rep
Table 4. LRP, line replacement parameters
LRP Image Line Replacement V6.0 EASI/PACE 08:29 17-Feb-97
FILE - Database File Name :pathname/filenlaps.pix
DBOC - Database Output Channel List } 2
DBOW - Database Output Window > 101
RMOD - Replacement Mode: ABOV/BELO/MEAN:MEAN
LINC - Line Increment Factor > 1
20 m from its true location. Although we
achieve the reported level of map accuracy, TM
imagery will always be printed at scales of
1:50,000 or smaller to avoid the blocky
appearance of the pixels.
Table 5. SMODEL procedure parameters
SMODEL Satellite Model Calculation V6.0 EASI/PACE 10:57 19-Feb-97
FILE - Database File Name :path/newfile.pix
DBGC - Database Ground Control Segment > 3
ORBIT - Orbit Segment Number > 2
MODEL - Satellite Model Segment >
MODINPUT- Modify Input :NO
ELLIPS - Ellipsoid for the Earth :E012
ERRUNIT - Error Unit: Pixel/Metre :METRE
REPORT - Report Mode: TERM/OFF/filename :path/smodel.rep
Table 6. Creation of new file for rectification in CIM2
CIM2 Create Database and Image Channel Files V6.0 EASI/PACE 14:14 16-Feb-97
FILE - Database File Name :path/newortfile.pix
TEX1 - Database Descriptive Text 1 :p18r39
TEX2 - Database Descriptive Text 2 :sortho file/northern Big Bend
DBSZ - Database Size: Pixels, Lines > 7556 5412
PXSZ - Pixel Ground-Size in Metres > 28.5 28.5
DBNC - No. of Channels: 8U,16S,16U,32R > 7
Table 7. Georeferencing segment set with GEOSET procedure in PCI
GEOSET Set Georeferencing Segment V6.0 EASI/PACE 14:10 19-Feb-97
FILE - Database File Name :path/newortfile.pix
UPLEFT - Upper Left Position for Database> 3317100 201900
LORIGHT - Lower Right Position for Databas> 3085020 447000
MAPUNITS- Map Units: PIXEL/UTM/others :UTM 17 R E012
Table 8. Parameters for satellite orthorectification procedure
SORTHO Satellite Image Orthorecification V6.0 EASI/PACE 14:22 16-Feb-97
FILI - Database Input File Name :path/nlapsfile.pix
FILO - Database Output File Name :path/newortfile.pix
FILEDEM - Database DEM File Name :
DBIC - Database Input Channel List } 1 2 3 4 5 6 7
DBOC - Database Output Channel List } 1 2 3 4 5 6 7
DBEC - Database Elevation Channel List >
BACKELEV- Background Elevation Value >
ESCALE - Elevation Scale and Offset > 28.5 0
DBIW - Database Input Window >
MODEL - Satellite Model Segment > 4
PXSZOUT - Output Pixel Ground-Size > 28.5 28.5
RESAMPLE- Resample Mode: NEAR/BILIN/CUBIC :NEAR
For example in a TM scene:
Region x offset (pixel) y-offset (pixel)
NW 1 0 0.5 -
NW 2 0.5 + 0.5 -
NW 3 1.0 + 0.5 -
NE 1 0 0.5 -
NE 2 0.5 + 0
NE 3 0.5 + 0.5 -
SE 1 0.5 + 1.0 -
SE 2 0.5 + 0
SE 3 0 0.5 -
SW 1 0.5 + 0.5 +
SW 2 0.5 + 0.5 +
SW 3 0.5 + 0.5 +
Table 9. Bitmap creation with THR
THR Thresholding Image to Bitmap V6.0 EASI/PACE 09:34 17-Feb-97
FILE - Database File Name :fltms/fileort.pix
DBIC - Database Input Channel List } 1
DBOB - Database Output Bitmap }
TVAL - Threshold Value (Min,Max) > 1 255
COMP - Complement: ON/OFF :OFF
DBSN - Database Segment Name :imageon
DBSD - Database Segment Descriptor :image bitmap excluding non-image border
Table 10. EASI procedure TMRAD
TMRAD - radiometric correction of TM imagery V5.3 EASI/PACE 09:10 15-Aug-95
FILE - Database File Name :/imagery/tm95/TM19950402
DBIC - Database Input Channel List > 1 2 3 4 5 6 7
DBOC - Database Output Channel List > 8 9 10 11 12 13 14
DATESITE - J.day, GMT, lat, long (W < 0) > 92 1511 28.8686 -82.4237
TMSLOPE - Slope for TM or MSS band > 0.0632 0.1254 0.0964 0.0907
0.0125 0.0055 0.0067
TMBIAS - TM or MSS offsets for channels 1> -0.118 -0.1935 -0.1697 -0.1628
-0.0248 0.1238 -0.0125
TMSOLAR - TM or MSS solar constants 1 - 7 > 195.7 182.9 155.7 104.7
21.93 1 7.452
SCALEF - Scale factor for units to counts> 500 500 500 500 500 100
500
TEXT - Text input :seg
Table 11. TMRAD output
Julian Day (year-day)= 92
Time (GMT) = 1511
Lat/long = 28.8686 -82.4237
*** Solar zenith angle = 40.5686 degrees ***
*** Solar azimuth angle = 119.377 degrees ***
*** Earth-sun distance = 0.999353 a.u. ***
cos(sza) = 0.759628
band # 1 10.00133387
slope bias solar irradiance refl/ct
0.0632 -0.118 195.7 148.852 0.00133387
band # 2 20.00283185
slope bias solar irradiance refl/ct
0.1254 -0.1935 182.9 139.116 0.00283185
band # 3 30.00255726
slope bias solar irradiance refl/ct
0.0964 -0.1697 155.7 118.427 0.00255726
band # 4 40.00357806
slope bias solar irradiance refl/ct
0.0907 -0.1628 104.7 79.636 0.00357806
band # 5 50.00235428
slope bias solar irradiance refl/ct
0.0125 -0.0248 21.93 16.6802 0.00235428
band # 6 6 0.0055
slope bias solar irradiance refl/ct
0.0055 0.1238 1 0.760612 0.0055
band # 7 70.00371355
slope bias solar irradiance refl/ct
0.0067 -0.0125 7.452 5.66808 0.00371355
Table 12. RAYRAD input
/pci/usgs/bin: rayrad
enter latitude (decimal degrees)
30.31
enter longitude (decimal degrees, where west < 0)
-83.92
enter time in GMT, (hhmm)
1537
enter Julian day
119
enter scalefactor (conversion of reflectance to counts for processed imagery)
500
Table 13. RAYRAD output
channel radiance reflectance counts
1 3.5128 0.0571 28.5
2 1.6372 0.0285 14.2
3 0.7739 0.0158 7.9
4 0.1991 0.0060 3.0
5 0.0026 0.0004 0.2
solar constants tau r tau g
196.0 0.17110 0.01
183.0 0.08920 0.03
156.0 0.04840 0.02
105.0 0.01850 0.02
21.7 0.00110 0.00
solar declination = 14.686262184445
solar zenith angle = 31.398820329709
solar azimuth = 113.20862349932
Table 14. Rayleigh subtraction model
doc *************************************************************
doc Correction of the radiance values gathered from using the
doc rayrad program. This program will allow corrections of the
doc digital numbers for bands 1 through 7
doc *************************************************************
doc_end
monitor = "ON"
report = "TERM"
file = "/path/fileort.pix"
see file
MODEL ON file
35 if (%%2 = 1) %8 = %8-29
rem 36 print "29 subtracted from dn for band 1"
40 if (%%2 = 1) %9 = %9-14
rem 41 print "14 subtracted from dn for band 2"
45 if (%%2 = 1) %10 = %10-8
rem 46 print " 8 subtracted from dn for band 3"
50 if (%%2 = 1) %11 = %11-3
rem 51 print " 3 subtracted from dn for band 4"
55 if (%%2 = 1) %12 = %12
rem 56 print " 0 subtracted from dn for band 5"
60 if (%%2 = 1) %13 = %13
rem 61 print " 0 subtracted from dn for band 6"
65 if (%%2 = 1) %14 = %14
rem 66 print " 0 subtracted from dn for band 7"
ENDMODEL
100 return
| Table 15. Darkwater values: |
|
Channel 1: 16 Channel 2: 12 Channel 3: 8 Channel 4: 6 |
Several indices are calculated at the final point in processing. An additional band is added to the .pix file for each index desired. Be sure to specify the correct band number when calculating an index. The examples below are based on the radiometrically enhanced data residing in bands 1-7. It is possible to protect previously processed image channels from accidental overwrites with the LOCK procedure.
2.4.1 Bitmap creation
A mask is employed to permit
processing and modeling of image-only pixels. .
The mask is created by selecting all pixels with
a value of > 0 in TM band 1 or 2. The new
bitmap segment is on at every image pixel and
off at all non-image pixels. It is important that
the operator check the band at full-resolution to
ensure there are no zero values in the image area
itself before running this procedure. The results
can be checked afterwards on full-resolution
display. Failure to check may result in missing
data in the index channels. The mask is created
by running THR as in Table 9.
The mask allows the operator to run processes on
image data only, maintaining values of zero in
non-image areas of the file.
2.4.2 Vegetation Index
The vegetation index, NDVI may be
calculated in the EASI procedure, RTR (Table
16). Alternatively, a model can be written to
take the input channels 3 and 4 and write output
NDVI to channel 8. Differences in NDVI
between two years may range from ~-90 to +90.
These values are re-scaled to range from 1 - 200.
2.4.3 Wetness Index
The wetness index may be calculated
with EASI procedure, ARI (Table 17). It may
also be run as an EASI model:
2.4.4 Thermal band 6 temperature
Temperature is calculated from TM
thermal band 6 (Markham and Barker, 1986) to
degrees Celcius plus 5. We add 5 degrees to adjust for
close to freezing temperatures which may
occasionally occur in the Big Bend region. The
model is run from a file as follows:
where segment 2 is a bitmap, masking image-
only pixels.
2.4.5 Water reflectance
Water reflectance can be calculated with
EASI procedure, ARI, (Table 18) or as follows:
2.4.6 Brightness
Brightness is merely TM band 2 with
the appropriate adjustments for radiometric
enhancement and atmospheric corrections. This
is sufficient for comparison of brightness between
scenes.
EQUATION 7. EASI MODEL FOR NDVI
%8 = 100 * (%4 - %3) / (%4 + %3)
EQUATION 8. EASI MODEL FOR WETNESS INDEX
%9 = %5 - %2
EQUATION 9. EASI MODEL FOR TEMPERATURE
IN CELSIUS
MODEL ON "/pathname/filename.pix
if %%2 = 1 %10 =
(1260.56/ln(60.776/(%6/100)+1)-268);
ENDMODEL
EQUATION 10. EASI MODEL FOR WATER
REFLECTANCE
%11 = %2 - %4
Table 16. Calculation of NDVI using EASI RTR
RTR Real Database Channel Ratioing V6.0 EASI/PACE 14:54 16-Feb-97
FILE - Database File Name :path/fileort.pix
CNUM - Channels for Ratio Numerator } 4 3
WNUM - Weights for Ratio Numerator > 1 -1
NCON - Constant for Ratio Numerator > 0
CDEN - Channels for Ratio Denominator } 4 3
WDEN - Weights for Ratio Denominator > 0.01 0.01
DCON - Constant for Ratio Denominator > 0
DBOC - Database Output Channel List > 8
SMOD - Scaling Mode: NONE/AUTO/LOGS :NONE
MASK - Area Mask (Window or Bitmap) > 2
ZERODIV - Value for Division by Zero > 255
Table 17. Calculation of the wetness index with EASI ARI
ARI Image Channel Arithmetic V6.0 EASI/PACE 14:34 16-Feb-97
FILE - Database File Name :path/fileort.pix
OPER - Operator: ADD/SUB/MUL/DIV/AND/OR:SUB
CNST - Input Scalar > 0
MASK - Area Mask (Window or Bitmap) > 2
DBIC - Database Input Channel List } 5 2
DBOC - Database Output Channel List > 9
ZERODIV - Value for Division by Zero > 255
AUTO - Autoscaling mode: ON/OFF/USER :OFF
RVAL - Function Min,Max, Output Min,Max>
Table 18. Calculation of water reflectance with EASI ARI procedure.
ARI Image Channel Arithmetic V6.0 EASI/PACE 14:34 16-Feb-97
FILE - Database File Name :path/fileort.pix
OPER - Operator: ADD/SUB/MUL/DIV/AND/OR:SUB
CNST - Input Scalar > 0
MASK - Area Mask (Window or Bitmap) > 2
DBIC - Database Input Channel List } 2 4
DBOC - Database Output Channel List > 11
ZERODIV - Value for Division by Zero > 255
AUTO - Autoscaling mode: ON/OFF/USER :OFF
RVAL - Function Min,Max, Output Min,Max>
Coastal and Marine Program >
Center for Coastal Geology >
Research by Theme >
Gulf of Mexico Tidal Wetlands >
Image Processing Methods - OFR 97-287 >
Chapter 2
U.S. Department of the Interior,
U.S. Geological Survey, Center for Coastal Geology
http://coastal.er.usgs.gov/wetlands/ofr97-287/chapter2.html
Address questions and comments to Trent Faust - Webmaster
Updated June 25, 1999 @ 10:31 AM
(THF)
