University of Southampton OCS (beta), CAA 2012

Font Size: 
Speeding up Georeferencing with subpixel accuracy
Gianluca Cantoro

Last modified: 2011-12-21


Geo-referencing a raster in a GIS environment is still a time-consuming operation and it usually requires a good level of patience and precision in choosing the right control points for the rectification.

When the geo-referencing has to be applied to a sequence of photos, for instance, from a flight sortie, the procedure, which is essential for the understanding of distances or landscape shape, takes a lot of time compared to the interpretation itself.

In this paper, a free software toolkit is presented which will allow any unskilled user to (partially) automate and speed-up the process of rectifying a raster in the real world coordinates with high level of accuracy and, most important, the possibility to edit and re-process the output for a further optimization.

What makes this software a unique and essential tool, is the fact that it does not provide the user with a distorted and geo-referenced image, instead it produces -in very short lapse of time- a list of matching points between an already positioned raster (e.g. a satellite image, a screen-shot from googlemaps or even just another raster from the same sortie) and the one that the user wants to geo-rectify (e.g. an oblique or vertical view of the same -or smaller- area as the previous input image or an overlapping image from the sortie).

This list of matching points, generated in no more than a couple of minutes by an improved version of a very complex (free) algorithm, is then converted into a list of ground control points to be loaded in QGIS or ArcGIS.

Since most of the times and depending on the quality of the images, the matching points are easily reaching a number of some thousands, a second script in the toolkit allows to filter this number to a more manageable one, by maximizing the minimum distance between each of the result of the process. In few seconds one can pass from a few thousands of points to one hundred homogeneously distributed points in the overlapping area between the two input rasters.

Lastly, since many (free or commercial) software packages allow the user to export the distorted images in a geotiff format, a third script in the toolkit helps the user to convert this (or whatever else) geotiff in a .jpg format accompanied by a worldfile. This combination of .jpg and .jpw files can then be used to geo-position the new raster in the same coordinate system.

The toolkit, written in Python and converted in windows binary to facilitate its use by the scientific community, is freely accessible through the website of FORTH-IMS ( The particular project was implemented under the initiatives of the ArchaeoLandscape European project (


gis; free software; georeferencing; rectification; photogrammetry