star mask for de-ringing in LR deconvolution

[lucchett]

Hi every body,
I am new to the Forum and the first impression is great.
After spending many days in trying to understand the deconvolution (R LR) process I am still far from solution.

I've read all the tutorials and tryied hundreds of combinations, but If I get a result is just by chance.

my problem is ringing (of stars) and the associated star mask.
I find very difficult to build a mask that's able to "protect" the small stars. I get hard ringing

The only way seems to use a PSF very small (about 1.0), but in my case should be more in the range of 1.5-1.8.  More, I think that using such a small PSF I get more artifacts than real details..


I've also tried to build the star mask in PS: I find easier to create a mask that protect small stars, but at the end it doesn't work. (opacity?)

I've read in another opic that the mask is not used to "protect" in case of deringing, but gives just a pattern to the algorithm.
Can someone explain, so that I can try to build the right mask?


Thanks in advance,

Andrea

[Juan Conejero]

Hi Andrea,

Building star masks is a complex task, but be sure we have the most powerful tools and techniques here; it's just that some practice (and a lot of patience!) is required.

In PixInsight, we have basically two different strategies to build a star mask:

- Use the StarMask tool. Optionally, in difficult cases you can use the HDRWaveletTransform tool before StarMask. This combo is extremely powerful. For example, we have managed to build a star mask to isolate all stars inside the core of M42/M43:

http://pixinsight.com/forum/viewtopic.php?p=2830#2830

- Use the ATrousWaveletTransform tool. In this case, the basic idea is to isolate the scales that define the stars that we want to protect. This approach can work very well in relatively simple cases. Consider this image:

http://forum-images.pixinsight.com/legacy/star-masks/atw/01.jpg

Here is an example of star mask built with wavelets:

http://forum-images.pixinsight.com/legacy/star-masks/atw/02.jpg

As you see, I have disabled large-scale wavelet layers where we have no object of interest, including the residual layer. The remaining layers, where stars are supported, have been biased to give them more relevance. The k-sigma noise thresholding parameters allow us to remove most nonstellar objects, as nebular features. This mask can be easily transformed with HistogramTransform and MorphologicalTransform to achieve the required star protection.

If you can upload an image where you are having trouble to build a star mask, I can try to build one as a small tutorial. This can help you (and others) in "getting the right direction" with these tools.