INP=(A,B,C),
where:
A is the first image to be correlated.
B is the second image to be
correlated [or a ground control
point file in graphics-2 format --
NOT CURRENTLY SUPPORTED].
C is an IBIS file containing the image
1 matching points in pixel (L,S) in
the first two columns. These points
can also be geographic if the first
image has a GeoTIFF label and the label
coordinate system is used (column 1
will be the East and column 2 will
be the North). The file must also
have 11 columns or more, and the
other 9 columns are filled (see main
help). The first two columns may
change slightly (see algoritm discussion).
OUT=D
D Is an image of the matching areas as
"chips". NOT CURRENTLY SUPPORTED.
ITIE=(X1,Y1,...,X3,Y3) This specifies three
control points in the first
input image to control
matching search and resampling
geometry. Can be omitted if
both images have GeoTIFF labels
and the GeoTIFF mapping is
desired for a matching model.
OTIE=(P1,Q1,...,P3,Q3) This specifies the correspon
ding control points in the
second image. If the second
input is a ground control
point file then these values
are latitude longitude pairs.
The number of values given to
OTIE must be the same as the
number to ITIE. Can be omitted if
both images have GeoTIFF labels
and the GeoTIFF mapping is
desired for a matching model.
MAGNIFY=(RL,RS) This specifies that sampling
from both images is magnified
by RL in the line direction
and RS in the samp direction
(default 1.0,1.0). The grid
is always FFTSIZE^2 so this
increases the correlation area
without increasing computation.
The magnification can be speci-
fied separately for line and
sample directions. If you use
these, you must specify both.
MAGSHRK="y" This specifies that the MAGNIFY
parameter will shrink by .9 for
every good fitting point. Also
only applies after fit to points
is obtained. When MAGMIN is
reached, then the shrink is no
longer applied.
MAGMIN=(RL,RS) This specifies the minimum vals
for MAGNIFY when MAGSHRK is
used. When MAGMIN is reached,
then the shrink is no longer
applied.
SEARCH=S This specifies that the search
area in the first input is
initially an SxS square re-
gion. Allow for uncertainty
and nonlinearity in the esti-
mation of matching location by
ITIE, OTIE and add 12 more to
S for ineffectiveness of cor-
relation near the window edge.
S must be less than or equal
to 128 which is also the
default.
MINSRCH=T This specifies that the
initial search S can be
reduced upon successful
matching to a minimum value of
T. Note that successful
matches refine the image to
image geometric model allowing
a reduction in search with a
consequent savings of
computation. The default
value is 50 and miminum value
is 32.
ZWIND=W The program calculates and outputs
the average brightness values
at the matching coordinates.
The average is calculated using
a W by W window. Default is 7.
ZREJECT=Z The average brightness is rejected
by setting the output Z value to
-999 if if falls below this value
The default is 5.
RETRY=(N,T,R) This instructs the program to
attempt more tries near a
location if a certain
correlation threshold is not
achieved. N is the number of
tries (1 to 5), with the first
try being the at the original point.
T is the correlation threshold
required to stop the retry at each
point. A failure number is
printed every time the threshold
test fails. If all retries for
a point fail then nothing is
written to the output files.
The retry pattern is to move
to the corners of a square
i.e. to move to (+/-R, +/-R) from
the original point.
'NOPRINT This specifies that the normal
printout is to be suppressed.
Error messages will still be
printed.
'NOHPF This specifies that the
default high pass filter in
the correlation is to be shut
off. The high pass filter
consists of zeroing the low-
order row and column of the
DFT matrices.
'NOSUBPIX This specifies that subpixel
correlation is not to be
calculated (see operation)
The correlation that results from a bad estimation of position will usually be not as good as one resulting from a good estimate. This parameter will recalculate the user-selected number of points over again at the end of processing. Usually, the predictor has optimized by the end of processing, so the results will be better.
"LINEAR" will use a 2-D linear fit to predict where a point in the first image will match in the second image. "QUAD" will use a 2-D quadratic fit, but only when enough points are available to solve for the coefficients of the quadratic. Three points are needed to solve for linear coefficients, and six are needed for quadratic. If the images have a keystone offset, use quadratic.
The value of this parameter (n) tells the routine to use the first n correlations to set the threshold. It will be the value for which 1/2 of n are above threshold and 1/2 n are below. No values are allowed into the fit until these n points determine the threshold. Then the convergence is held for the next set of n points while the fit is being refined. This whole procedure is to prevent a misleading fit on the first point from sending the whole fitting process in the wrong direction. The user provided fitmin will be ignored, it is set by the process just outlined.
0 = default, no retries when zerolim exceeded 1 = one retry at half size 2 = one retry at half size, another retry at quarter size This allows correlating closer to the image edge. The correlation threshold is toughened so that the correlation must be at the automatic threshold instead of within .5 of it.
Setting this to a value limits the number of zeros in either image chip to that percentage of zeros. For the first image, exceeding this percentage will cause a message "point outside first image" and the point will not be used (column 9 will have -9999). For the second image, which may be searched, the individual chips are rejected with no comment. If the second image is 50% zeros a message "point outside second image" is given. Parts of a chip cut off by an image edge are treated as zeros. The parameter ZEROLIM2 allows the second image value to be set independantly. If it is not used, then the default value of -1 causes the ZEROLIM for the second image to be the same as the first image.
See routine piclsqm2 for use of these parameters
tilts initial match predictor by this angle in degrees. Gives better correlation if the tilt is accurate. After the solved fit for a match predictor, this angle will no longer be used or have any effect on results.