IMCORR is a package distributed by the National Snow and Ice Data Center (did you know we have one of those?), or NSIDC, that performs image cross correlation between two images using a comparison between a moving image chip in each image. It measures the displacement (in pixels) between the objects found in the two images, and writes that out to a text file. It is used to calculate how quickly glaciers are moving by calculating displacements. If you have an image of a glacier from two different time periods, this little program will calculate how far crevasse zones and other discernible features have moved between two images. I know I know. I have been talking about trees, and now ice. Bear with me.
IMCORR can also be used for simple image georeferencing, say when we don’t want to collect our own tie points. We can look for correlations between two images, and tabulate those as points, use them with a thin plate spline, in say gdalwarp, and thus correct for distortions in one dataset with another one.
Now, this is all in theory, and I haven’t tested it yet. But I’m thinking this may be a really useful technique, especially because we already have well referenced high-resolution aerials for so many areas. I have an effective 3″ color infrared (CIR) dataset for an entire county kicking around somewhere that has been referenced to frame on center, but with no terrain correction or other optical distortion corrections applied. I’m hoping to cross correlate these with a 6″ georeferenced aerial in order to make these data useable. 3″ CIR could be very useful for all sorts of fun applications… .