Does cropping and adding an alpha layer to a raster map yield better georeference results?

Question

I have a collection of nine high-resolution map scans illustrating former borders between two nations. The scale is about 1:13,265 for each of the nine images. I want to georeference these maps using QGIS so that I can trace the historical borders, but I want to know whether processing the images - in GIMP, for example - will influence the accuracy of the results.

Right now, each map is on a rectangular sheet with white background. If I crop everything outside the frontier zone and replace the cropped out space with an alpha layer, would this image produce more accurate, less distorted results than the rectangular originals? Using the same ground control points for any given transformation type (e.g., linear, polynomial-1, thin plate spline), would it yield larger or smaller residuals?

Answer 1

If you have a very large map and you want to georeference only a small portion of it, it would make sense to crop it, as you will have less to georeference, less error, less area. For example, if one would have a world map and needs only New Zealand, to georeference it you would have to place points on all the map; from my point of view it makes sens to crop the unwanted bits, as long as it would not influence the end result. Also cropping a large map, would help on the size part, meaning it's easier for ArcMap(QGIS) to draw a 100 Mb file, then an 1 Gb file.

I don't see how would an alpha channel improve your accuracy, when it only has a role in transparency.

So to conclude, yes on cropping, it would make your file size smaller and easier to handle, alpha has no influence in gereoferencing afaik, try to place your points as accurate as possible, try all the transformation before settling for the one that fits your needs the best.

Here is a great post with some good practical advice on georeferencing.