Lanczos Resampling Issue

[akulapanam]

In order to register my osc ha frames to my rgb frames I have been using the Lanczos resampling method. Unfortunately last night I realized that this was causing extensive ringing and in general noise. I tried changing the clamp with out effect and then tried a couple other methods. The good news is the other methods resolved the issue but the bad news is the overall brightness and contrast of the image decreased dramatically. Any ideas on how to get the best of both.

[Juan Conejero]

Interpolation algorithms and their implementations in PixInsight are reasonably documented:

http://pixinsight.com/doc/docs/InterpolationAlgorithms/InterpolationAlgorithms.html

This documentation is a bit outdated (we now have better implementations), but the essential information is there. You are facing the eternal dilemma between ringing and aliasing. Each data set poses completely different problems and requires different solutions.

Just a note about a common caveat: if you are tempted to use the bicubic B-spline algorithm, be advised that this is probably the worst choice. Bicubic B-spline implements an old resampling method by Paul Bourke and is being kept in PixInsight for historical and compatibility reasons. This algorithm will be removed in a future version. Also try to avoid bilinear interpolation, as it is the worst option in terms of aliasing.

[akulapanam]

What about the brightness vs contrast issue?  Also what would you recommend instead of lanczos?

[akulapanam]

Version A used Lanczos in PixInsight to Resample Ha to RGB- best stars but lost blue nebula, very grainy, stars have severe ringing at 100%. Deconvolution and HDR can't be used because of grain

http://www.astrobin.com/199104/0/
 
Version B used B-Spline in PixInsight to Resample Ha to RGB

http://www.astrobin.com/199104/B/
 
Version C used B-Spline in another program to Resample Ha to RGB

http://www.astrobin.com/199104/C/

[akulapanam]

Here is a much better comparison. I cropped each to IC1284 itself at 100% zoom.  The original image is close to 9mp.

http://www.astrobin.com/full/199142/0/

[Juan Conejero]

So you've found "another program" that works a treat. Congratulations, but this forum is for "this program", not for "another program". The documentation linked above describes ringing and the options you have to prevent it in PixInsight. Bicubic spline interpolation (not to be confused with bicubic B-spline) is one of them when the source data is too weak for Lanczos. If you don't like our implementations, or if you think they don't work for you, then use "another program". Just bear in mind that we are not really interested in "another program" here, just "this program".

[akulapanam]

That really was not my point. I think the b-spline in PixInsight produces acceptable results although I'm not sure why it changes brightness/contrast.   The other program lanczos has the same issue has pixinsight.  Right now I'm trying the PixInsight drizzle approach and we will see what happens.

[akulapanam]

So here is the version of the image using drizzle.  Not too bad but pretty grainy.  I had to dump on noise reduction.  Also for some odd reason my drizzle stack ended up a little left which is why part of the right is cropped off.

http://www.astrobin.com/199104/D/