Hi Georg,
I am not sure if this will help.
It will improve PixInsight's image registration capabilities considerably. As you know well, we currently have a serious limitation in StarAlignment due to the use of triangle similarity for the star matching routine. This basically limits SA's capabilities to affine transformations, plus a small amount of global distortion thanks to the implemented optimization techniques. The use of quad descriptors will allow us to handle projective transformations (homographies).
Note that it is just the kind of device used to feed the pattern matching algorithm what changes; the rest of StarAlignment remains the same. Since quads can be characterized with only two coordinates, we can reuse the whole pattern matching routine exactly as it is now. Instead of working in a triangle space (where the coordinates are the relations between triangle sides), we work with pairs of quad local coordinates. We continue working in a two-dimensional space.
I don't know if the new SA tool will be able to align all the problematic wide field images that we have seen here; it probably won't, since a homography cannot model arbitrary local distortions. However, the new tool will be much more flexible and will lead to much more accurate results under difficult conditions.
