Trying to register image on itself fails...

[georg.viehoever]

... see attached log.
Georg

Code: [Select]

...
Registering target image 11 of 38
Loading target file:
C:/Users/Public/Pictures/Astro/2011_10_25_andromeda_85mm/fit/lightcalibrated/debayered/debayer_cal_light20_c.fit
Reading FITS: 32-bit floating point, 3 channel(s), 3908x2602 pixels: 100%
Structure map: 100%
Detecting stars: 100%
11655 stars found.
Matching stars ...
* Reference image: Limiting to 2000 brightest stars.
* Target image: Limiting to 2000 brightest stars.
2000 putative star pair matches.
Performing RANSAC ...
2000 star pair matches in 1 RANSAC iterations.
Summary of model properties:
Inliers     : 1.000
Overlapping : 1.000
Regularity  : 0.995
Quality     : 0.999
Root mean square error:
?RMS  :  0.000 px
Average RMS error deviation:
?RMS  :  0.000 px
Peak errors:
?xmax :  0.000 px
?ymax :  0.000 px
Transformation matrix:
     +1.0000     +0.0000     -0.0000
     +0.0000     +1.0000     +0.0000
     +0.0000     +0.0000     +1.0000
scale    : 1.000
rotation :     +0.00°
dx       :     -0.00 px
dy       :     +0.00 px
*** Error: Singular matrix in 2-D surface spline generator

*** Applying error policy: Continue on error.
...


[Juan Conejero]

Hi Georg,

This was an old bug that I fixed some versions ago; now StarAlignment can register an image on itself without problems (just try with a different image). However, the 2-D surface spline generation routine can fail sometimes with a 'singular matrix' exception. This error is more probable for more ill-posed problems such as autoregistration.

While I fix the singular matrix problem (which is not easy BTW), just don't use surface splines and there should be no problems. Surface splines are only necessary for mosaic construction.

[georg.viehoever]

I think it should be sufficient to capture the case where the transformation matrix is sufficiently close to
     +1.0000     +0.0000     -0.0000
     +0.0000     +1.0000     +0.0000
     +0.0000     +0.0000     +1.0000
All other cases of singular spline matrices should be rather irrelevant.

Georg

[Juan Conejero]

Actually, the singular matrix problem may arise in many other cases; the autoregistration case is marginal. See for example:

http://pixinsight.com/forum/index.php?topic=3685.0