GradientMergeMosaic odd results

[pfile]

i've never tried to upload a fits file to astrometry... there is of course no way to specify the origin of a fits file (meaning, which corner of the image is byte 0), so maybe that's what astrometry was complaining about. sounds consistent with the flipping you observed.

anyway glad to hear it's working right. GMM is really sensitive to brightness differences at the overlaps and so i guess that 'extra' stuff was really messing it up.

a blind solver would be good - i think the astrometry.net source is open source so there's something to build on. i remember hearing that the algorithm shares some similarities with how shazam/soundhound identify music.

rob

[Cosmick]

a blind solver would be good - i think the astrometry.net source is open source so there's something to build on. i remember hearing that the algorithm shares some similarities with how shazam/soundhound identify music.

A way to use ANSVR would be nice as there are already several programs connecting to this for blind solving. This saves a lot of space as there is only one set of index files needed.

Application developers: information about adding ansvr communication to your software application is here: ansvr developer notes.

[gnewell]

i've never tried to upload a fits file to astrometry... there is of course no way to specify the origin of a fits file (meaning, which corner of the image is byte 0), so maybe that's what astrometry was complaining about. sounds consistent with the flipping you observed.

It was PI that complained. Not Astrometry.

[gnewell]

@cosmik,

On issue is that ansvr vs. nova.astromety.net is they seem to be on different versions. I don't recall the details but if memory serves the API and file support are different.

Speaking of file support, xisf support in both of those would be great.

[Andres.Pozo]

a blind solver would be good - i think the astrometry.net source is open source so there's something to build on. i remember hearing that the algorithm shares some similarities with how shazam/soundhound identify music.

Astrometry.net is a blind solver, but ImageSolver is much more precise.
Watch this example of the same image solved with ImageSolver with distortion correction and astrometry.net.

[pfile]

yes i agree, i don't think i would use the astrometry code for anything but a starting solution. ImageSolver is way more precise.

rob

[Cosmick]

yes i agree, i don't think i would use the astrometry code for anything but a starting solution. ImageSolver is way more precise.

rob

I agree as well. I often get files from other people that have no acquisition, position or object information and a blind solver would save a lot of trial and error trying to find the right parameters to feed to ImageSolver.
I also forgot that ANSVR is for Windows only so would not work as a solution for Mac and Linux.

[pfile]

it would work - ANSVR is just a wrapper for astrometry.net, and astrometry.net can be compiled for any platform. ANSVR just duplicates nova.astrometry.net 's HTTP api for local use. PI could probably just call the command-line version of ANSVR to get a solution rather than using http post commands.

ANSVR is built on top of Cygwin which implies it could be ported back to linux or osx, if you wanted to take the http route.

rob