Hi Wade,
Of course, I am very interested in taking a look at those images. You can upload them to the ftp server, as usual (zipped, please don't use rar as I can't open the files). I am sure this will lead to an improvement, as always happens with your images
Thank you!
StarAlignment's star matching algorithm is based on triangle similarity. Triangle similarity is invariant to translation, rotation, mirroring and scale change. Note that this covers an affine transformation only partially (since shearing breaks triangle similarity). In addition, I implemented a triangle selection strategy that is much more tolerant to small-scale distortions than the original algorithms (please refer to
this thread and skip to the
Technical Description and Reference Information section for details). However, large-scale distortion is incompatible with this star matching scheme.
To perform the final image registration, StarAlignment fits a homographic transformation (an eight-parameter transformation, providing for full affine transformations plus chirping and keystoning). However, the current star matching algorithm imposes the limits that I have described above. A homographic projection allows for much more flexible registration tasks in future versions of SA and derived tools, which must implement more sophisticated feature matching algorithms, including non-star-based and non-astronomical applications, as lunar mosaic and panorama generation. In other words, SA is a powerful platform for development of image registration systems; the current SA tool is just an initial stage.
