Author Topic: ABE polynomial fit  (Read 2353 times)

Offline fjabet

  • Newcomer
  • Posts: 6
ABE polynomial fit
« on: 2015 December 13 03:32:52 »
Hello,

I would like to know more details on the ABE polynomial fit. Does it use a polar or cartesian system (polar would make sense actually) ?
Is the fit a global n degrees over the whole surface, or is it more complex than  this ?
Is it a straight forward Laplace fit or does it use slopes as well (Hermite).
I've seen Runge bouncing using high orders, so I guess it is not using the derivate ?
I understood that the DBE uses a spline fit. Is it a pure spline or something more complex ?
Is there an detailed  documentation of ABE and DBE ?

Thanks in advance.

Best regards,

Frédéric.

Offline fjabet

  • Newcomer
  • Posts: 6
Re: ABE polynomial fit
« Reply #1 on: 2015 December 16 00:07:39 »
I up this question, I'm preparing an advanced training for french AIP and I would need some clarifications. Is there a support contact address from Pleiades ?

Offline Juan Conejero

  • PTeam Member
  • PixInsight Jedi Grand Master
  • ********
  • Posts: 7111
    • http://pixinsight.com/
Re: ABE polynomial fit
« Reply #2 on: 2015 December 16 00:37:27 »
I can answer the DBE question for sure, since I designed and wrote that tool. DBE uses two-dimensional smoothing surface splines, also known as thin plates. Second-order splines are used by default (that is, the generated interpolation device guarantees continuity of second derivatives by default). Our implementation of thin plates on the PixInsight/PCL platform is based on the solution given in the 1996 book Numerical Algorithms with C by Giesela Engeln-Müllges and Frank Uhlig. I have used different adaptations of the same algorithms in other PixInsight tools such as StarAlignment and DynamicAlignment.

The AutomaticBackgroundExtractor (ABE) tool was designed and initially implemented by Carlos Milovic, so you better ask him for details. AFAIK, ABE uses a global 2-D polynomial interpolation in rectangular coordinates. I think it does not have derivative continuity constraints, but I may be wrong. This interpolation scheme, as all global (that is, non-piecewise) polynomial-based interpolations, tends to generate ripple artifacts using high orders. However, polynomials are much more rigid than splines, which is potentially more robust in some cases, especially where adaptation to local variations is not a desirable property of the solution. Each scheme has its own strong and weak points.
Juan Conejero
PixInsight Development Team
http://pixinsight.com/