I want to test if two raster images for the same scene taken in different moments are aligned. I would like to do this automatically as I will be working with a lot of scene pairs. The process would not need to be perfect I would like to spot obvious miss-alignments (i.e. more than 2 pixels) and flag them for manual inspection.

Is there a standard way to do so?

I am using python + GDAL

  • I am using python + GDAL – amaurs Jan 6 '17 at 22:01
  • 1
    gdalinfo should give you, among many other things, the origin, cell size and extent (corner coordinates). If those match, your rasters are aligned. – J Kelly Jan 7 '17 at 3:39
  • @JKelly That is not necessary true. In satellites like Rapideye, scenes can carry errors even when the metadata is correct. This can be spotted manually using Control Ground Points, however I am dealing with many images (around 8000+) so it would be desirable to have a first filter to spot these errors. – amaurs Jan 9 '17 at 18:25

It sounds like you are looking for automated image registration.

I would use something like ENVI (Automatic Image to Image Registration) or ERDAS Imagine (Autosync) to perform tie point generation which you can then inspect.

For an opensource option, check out RSGISLib which is available as a package via conda (Win64, OSX and Linux)

The GDAL project has this functionality as well, but it only seems to be exposed via the C++ API, not via a command line tool or the python bindings.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.