PixInsight Forum

PixInsight => Wish List => Topic started by: fneyer on 2012 April 13 00:36:34

Title: Extend LinearFit capabilities
Post by: fneyer on 2012 April 13 00:36:34
A great improvement to the existing LinearFit process would be the ability to automatically detect overlapping regions of two images. This region then will be used to estimate the parameters of the linear function. Once estimated, the function is applied to the whole image.

Instead of an automatic detection of the overlapping regions, the area used for parameter estimation could also be specified by a preview on the reference image.

Title: Re: Extend LinearFit capabilities
Post by: Juan Conejero on 2012 April 17 04:17:32
Hi Fabian,

The LinearFit algorithm requires that both images be previously co-registered, so in general you should use StarAlignment before LinearFit. LinearFit will reject all pixels outside the intersection between both images, since it will reject all zeros automatically.

An important design principle of PixInsight is modularity: Instead of reimplementing the same functionalities in all tools requiring them, try to separate the different tasks and isolate them into smaller and more efficient tools. In this case, StarAlignment provides a frame adaptation feature for convenience (it is essentially the same algorithm implemented as LinearFit) because it is a very small and relatively simple task. However, reimplementing StarAlignment's functionality in LinearFit, even if only partially, doesn't seem like a good design decision.

Title: Re: Extend LinearFit capabilities
Post by: fneyer on 2012 April 17 07:08:29
Thanks Juan, I understand your principle and it makes sense.
How about the preview box used for parameter estimation? This would not be in conflict with SA and probably is not a big deal to implement. I would appreciate to have this flexibility.