I'm working on Astrometry.net "picture-in-picture" mode for Julianne Dalcanton. Given a large (angular scale) image and a set of small (angular extent) images that are known to be within the large image, first nail the large image to a global astrometric frame, then nail the small images within the big one. Then tune up.
In her case, the target is NGC300, and the sample images she sent are from DSS (30x30 arcmin) and HST (~3x3 arcmin): three drizzled images.
It turns out that the HST images have astrometry good to a pixel or two, but subpixel is required.
Here is the overlapping region between the inner two HST images:
To me it doesn't look like the big image is going to help nail down the astrometry of the small images at all... but we'll see.
Here's a closer look at the intersection, with one image plotted in red, the other in blue, and the sources marked. There's clearly a shift, though I'm not ready to attribute that to the images rather than my plotting code -- lots of opportunities for off-by-one FITS pixel-indexing errors! (The blue squares are a plotting artifact.)
