Shift dim in x and y

[pcooley]

It is surprising to ask a question about the simplest task I’ve put to the SNAP program yet! I have slight offsets between SAR images and simply want to shift each of them in x and y to match a master vector dataset. These are small offsets in the existing projection, on the order of 20 m or less, before processing with other techniques. Your thoughts are welcome and appreciated for such a simple task!

[marpet]

There is no shift operation available in SNAP.
But you could create a subset and cut-off some pixels.
The question is also is the shift constant over the whole image?
Maybe it is better to improve the geo-location?
You could use GCPs for this. I’ve mentioned it here:

Manually georeference offset imagery in SNAP

hello Chui, The first thing you can try is to set the property s3tbx.reader.olci.pixelGeoCoding. Either by setting the property in a text file, as described here: Or in the GUI (Tools / Options): [image] This will improve the geo-coding if you still use the tie-points as basis. Next version of SNAP/S3TBX will have this setting as default. If this is still not sufficient. you can set ground control points (GCPs) in the scenes for known locations. Afterwards you can create a new geo-codi…

Have you already performed the usual processing steps for SAR data?

[pcooley]

Thank you Marco. Yes, we are dealing with the S1 timeseries where the geolocation is plus or minus a pixel or so on some images. In our timeseries the offset in some images is only on the Y axis so there is not a need to remap all the pixels using a GCP model. Perhaps this is the only way to do this in SNAP. We are extracting pixel data from a rather large set of images so these offsets among images contain backscatter from sliver/edge targets we do not wish to describe statistically. Can the metadata in the *.dim file be changed in a manner like that can be done with a geotiff world file, for example? The approach I hope to employ does not change the array but instead only the geocoding of it. I have a programmer that could manipulate the dim file, if that is a viable approach and if you could point us to the right metadata records. Your thoughts?

[pcooley]

Sorry…missed your last question. The images are sigma naught in dB using meters (UTM N WGS84).

[marpet]

Modifying the *.dim file would work.
Having the data reprojected to UTM you can change the values of the affine transformation.

In the *.dim file you will find a section (new line 40) like this:

    <Geoposition>
        <IMAGE_TO_MODEL_TRANSFORM>3.4521084625273746,0.0,0.0,-3.4521084625273746,471058.4096426825,7795161.477930118</IMAGE_TO_MODEL_TRANSFORM>
    </Geoposition>

These are the values for the matrix in the order [m00, m10, m01,m11,m02,m12]

      [ x']   [  m00  m01  m02  ] [ x ]   [ m00x + m01y + m02 ]
      [ y'] = [  m10  m11  m12  ] [ y ] = [ m10x + m11y + m12 ]
      [ 1 ]   [   0    0    1   ] [ 1 ]   [         1         ]

m00 and m11 specify the pixel spacing in x and y and m02 and m12 specify the upper-left corner.
You need to adjust m02 and m12 to shift image

[pcooley]

Great…thank you!