ANSI-C program: init_offsetm.c
NAME
init_offsetm - initial registration offset estimation for
multilook intensity images.
SYNOPSIS
init_offsetm <MLI-1> <MLI-2>
<DIFF_par> [rlks] [azlks] [rpos] [azpos] [offr] [offaz]
[thres] [patch] [cflag]
| <MLI-1> | (input) intensity image 1 (float) (reference) |
| <MLI-2> | (input) intensity image 2 (float) |
| <DIFF_par> | (input/output) DIFF parameter file |
| [rlks] | number of range looks (enter - for default: 1) |
| [azlks] | number of azimuth looks (enter - for default: 1) |
| [rpos] | center of region for comparison in range (enter - for default: image center) |
| [azpos] | center of region for comparison in azimuth (enter - for default: image center) |
| [offr] | initial range offset (enter - for default from DIFF_par) |
| [offaz] | initial azimuth offset (enter - for default from DIFF_par) |
| [thres] | correlation SNR threshold (enter - for default, default: 7.000) |
| [patch] | correlation patch size (enter - for default 128,256,512,1024, default: 1024) |
| [cflag] | copy offsets to the range and azimuth offset polynomials in DIFF_par (default=0:no 1:yes) |
EXAMPLE
init_offsetm 19990421.mli 19990422.mli
19990421_19990422.diff_par 3 3
Determines an initial registration offset (in slant range and azimuth) between the two indicated multilook intensity images 19990421.mli and 19990422.mli
DESCRIPTION
init_offsetm reads two
(multi-look) intensity images and the corresponding DIFF
parameter file. For a multi-look intensity image at a user
defined location, the two-dimensional cross correlation function
is calculated and the peak of the correlation function yields an
initial estimate of the offset between the images in slant
range and azimuth.
Confidence in the offset estimate is measured by comparing the height of the correlation peak relative to the average level of the correlation function to yield a correlation SNR. SNR values greater than 7.0 indicate registration estimates better than 0.5 pixel.
Due to low contrast some additional multi-looking of the images can improve the correlation between the images with some loss of resolution. Generally operating with rlks=1 and azlks = 1 is recommended.
In some cases, when images have very few features, it may become difficult to find a region where a reliable offset can be measured. In this case, examine the image (for example with dispwr), and locate an area where there are discernible features such as lakes or mountains. Then note the range and azimuth coordinates. These coordinates can then be entered as the rpos and azpos command line parameters so that the offset is estimated in the region containing the visible features. In the case of a relatively small feature, reduce the size of the patch to cover the area containing the feature. The patch size must be large enough to however take into account the range and azimuth offset between the 2 images.
OPTIONS
To optimize, both, robustness of the estimation as well as offset
estimation accuracy, init_offsetm may be applied
iteratively, starting with stronger multi-looking (such as 3x3)
with consecutive reduction of the multi-looking to the full
resolution (i.e. 1x1).
In order to refine a geocoding lookup table, the offsets of a transformed image can be determined and used to refine the geocoding lookup table (program gc_map_fine). This is usually a simulated SAR intensity image as created with the program geocode or pixel_area.
In measuring the offset, the transformed image (i.e. for example the simulated SAR image)should be taken as the reference geometry, in order to assure that the correct sign is used in the correction of the fine registration offsets. The user also has the option to copy the range and azimuth offsets to the offset polynomial constant terms. This is useful when performing a geocoding refinement and there are insufficient areas in the scene to generate an array of offsets. The user should select the rpos and azpos parameters to match an area of the scene with contrast. Generally, the constant range and azimuth correction for the geocoding refinement is adequate.
In cases where there are few features in the image, the user may select a smaller patch size using the patch command line parameter. Valid values for the the patch size are 128,256,512, and 1024.
SEE ALSO
typedef_ISP.h, typedef_DIFF.h .