New Algorithms for DBE

[frasax]

Hello,

this is not so much a feature request but rather a cue to some interesting algorithms that could be used to improve the DBE calculations.
In an open source project that I am involved with (have to say I am not a programmer) we used beside spline algorithms such as RBF and Kriging to calculate the background models. Especially the RBF (Radial Basis Function) leads to very good and fast results. Kriging is superior in some special cases.

Releases · Steffenhir/GraXpert

GraXpert is an astronomical image processing program for extracting and removing gradients from the background of your astrophotos. - Steffenhir/GraXpert

github.com

This is the link to our open source project. We would be very glad if the programmers of PixInsight would consider to use those algorithms in DBE or ABE. Problem might be that all the packages we used are coded in Python. But the RBF algorithm was already transferred to C in another image processing software.

For more Infos on RBF see

scipy.interpolate.Rbf — SciPy v1.13.0 Manual

docs.scipy.org

For more Infos on kriging see

pykrige.ok.OrdinaryKriging — PyKrige 1.7.1 documentation

geostat-framework.readthedocs.io

CS Frank

 

[Juan Conejero]

Hi Frank,

Thank you for the suggestion. We already have several radial basis functions implemented in our code base (variable order, thin plate spline, Gaussian, multiquadric, etc), along with the associated surface interpolation algorithms. The thin plate spline is just a particular case of RBF with remarkably good properties, which we use for many applications in PixInsight, including astrometric solutions, image registration (both tasks with arbitrary distortion support) and background model generation, among many others. The current DBE tool was tagged as obsolete several years ago. A completely new tool (DynamicBackground, DB) is already in the design phase and will be implemented as soon as possible; unfortunately, I have been forced to postpone it in several occasions because of other high priority tasks. I already know kriging and would like to implement it in the future, time permitting (time and human resources are always the main limiting factors for us).

Thanks for the information on your open source project, which looks great indeed. Unfortunately, any software project (be it open or closed source, free or commercial) with applications in the fields of astronomy and/or image processing is a competitor for us because it has the potential to hurt our chances of selling licenses.

 

[frasax]

Hi Frank,

Thank you for the suggestion. We already have several radial basis functions implemented in our code base (variable order, thin plate spline, Gaussian, multiquadric, etc), along with the associated surface interpolation algorithms. The thin plate spline is just a particular case of RBF with remarkably good properties, which we use for many applications in PixInsight, including astrometric solutions, image registration (both tasks with arbitrary distortion support) and background model generation, among many others. The current DBE tool was tagged as obsolete several years ago. A completely new tool (DynamicBackground, DB) is already in the design phase and will be implemented as soon as possible; unfortunately, I have been forced to postpone it in several occasions because of other high priority tasks. I already know kriging and would like to implement it in the future, time permitting (time and human resources are always the main limiting factors for us).

Thanks for the information on your open source project, which looks great indeed. Unfortunately, any software project (be it open or closed source, free or commercial) with applications in the fields of astronomy and/or image processing is a competitor for us because it has the potential to hurt our chances of selling licenses.

Hello Juan,

thanks for your reply. I remember that you mentioned some time ago that you plan to rebuild the DBE process.
With our project, I was able to quickly and easily achieve results that would have taken me more time in DBE. However, there are cases where DBE is the better choice because of the better setting options.
I see my post as info only. We also have no interest in competing with PixInsight. On the contrary, we would be happy if our project could contribute in any way to improve PixInsight. Our users are predominantly PixInsight users who will continue to use PixInsight.
In my opinion, PixInsight has no serious competitor.

CS Frank

 

[Juan Conejero]

We also have no interest in competing with PixInsight. On the contrary, we would be happy if our project could contribute in any way to improve PixInsight.

Excellent. In such case I invite you to contribute to our code base instead of working on a parallel project. Our C++ development framework is open-source under a liberal BSD-like license. In this way you would be contributing to make PixInsight better and stronger.

By developing a parallel project you are going to compete with us, and we'll be forced to fight against you, sooner or later, with all of the resources we are capable of. This is inevitable because you are providing reasons to stop using PixInsight and use a freeware application instead. We cannot compete with freeware applications in terms of price, and price is, unfortunately, one of the strongest motivations that drive potential users. On the other hand, our human resources are limited and our scope of work is very large. This means that a small group of skilled developers focused on a very specific and important topic (such as background modeling) can hurt us significantly. This has happened in several occasions, and has happened recently, forcing us to work over our capabilities for too much time and putting the entire PixInsight project under risk.

In my opinion, PixInsight has no serious competitor.

Everything looks different when you have to pay salaries, working places and their maintenance, development tools and resources, servers and their maintenance, content delivery network services, machines, user support resources, etc, etc... and you have invested everything. PixInsight is much more fragile and vulnerable than most of our users think. The line that separates us from a scenario where PixInsight is inviable is thin, although we try to strengthen it every day, of course

 

[frasax]

xcellent. In such case I invite you to contribute to our code base instead of working on a parallel project. Our C++ development framework is open-source under a liberal BSD-like license. In this way you would be contributing to make PixInsight better and stronger.

Hello Juan,

GraXpert is a Software that exists along with PixInsight and all the other Software out there. Cause not every beginner will buy a license of whatever software. And beginners are important, because they become pros later.
For our programmers it would be a lot of work to learn all the functions behind the PixInsight Development Framework. Also, they are reporting that they cant find the DBE within the modules anyway.
On the contrary it would be easy for you to use the algorithms we use and just give it a try. My experiences with our software are in a nutshell: Its easier to use then DBE (just a few parameters) and the results are as good if not better (in some cases) than DBE. That would be beneficial because what is annoying to new PixInsight users is its complexity.

Cheers
Frank

 

[Juan Conejero]

Our BackgroundModelization module was not released as an open-source product many years ago for strategic reasons. Currently we could release it without major problems, but we consider it obsolete code. The new DB process will be released as open source.

Most of the algorithmic infrastructure used by the DBE process is available in our PCL library. I am surprised by your comment about the amount of work required to contribute to our C++ development framework, since it is quite well structured, well documented, and there are lots of examples that can be used to start being productive quickly.

Thanks for the offer. We generally prefer to use our own resources and R&D work instead of depending on third-party products.

My experiences with our software are in a nutshell: Its easier to use then DBE (just a few parameters) and the results are as good if not better (in some cases) than DBE.

Thank you for the heads-up. In such case we'll have to work hard to provide a much better tool. As for complexity, you surely know that complexity is one of our main design principles, so maybe you can have an opportunity in this sense.