StarMask - dealing with large star

[joelshort]

I'm trying to create a star mask and am running into a problem with a large bright star.  The bright star doesn't get picked up by the starmask script. If I increase the scale to the point where it does pick up the bright star, the other stars of the mask are too large and I end up with halos around all the other stars. 
Attached is the image I'm trying to create a star mask from.  It's the central bright star that doesn't get picked up. 

What I'm seeking to do is apply the stars from a MaskedStretch image onto this image using a star mask, to maintain the contrast of the original stretched image  and add the stars from the MaskedStretch image.

[MortenBalling]

Hi Joel

One "cowboy trick", that I've used in situations like this, is to clone the image I'm working with. Then I resample the cloned image to 50%, and use the smaller downsampled image to make my starmask. You might also want to try to lower the values for Structure Growth and especially Smoothness. Then you get square stars in the mask, but "no free lunch, and minor problem". The downsampling makes the bigger stars easier to handle, and also speed up the whole trial and error process (by a factor of four! ).

Once you have a good starmask, you resample that back to the size of the original image (using resample = 200%). Be aware that PI can't work with half pixels, so if the original image had odd numbers of pixels in one or both dimensions, for example 2997x1999 pixels, you will have to manually enter the resolution of the original size, when upsampling the starmask.

Br

Morten

[joelshort]

Thanks Morten, I'll give that a try.

[RickS]

I often process the really big stars separately, building masks for them with RangeSelection, PixelMath and other tools.  It can be fiddly but worthwhile.

Cheers,
Rick.

[MortenBalling]

@Rick

Wavelet Transform can also be used, disabling the R layer, together with convolution and Pixelmath.

Br

Morten

[RickS]

Wavelet Transform can also be used, disabling the R layer, together with convolution and Pixelmath.

Thanks, Morten.  I have used wavelet operations to build star masks before but not tried to do it for large stars only.  I'll have to give it a try.  I wonder whether discarding both the smallest scales and the residual would work well to build a mask for larger stars?

Cheers,
Rick.

[joelshort]

Well guys, now you're talking way over my newbie head.  Is there a tutorial you can point me to that discusses building star masks with these other tools? 

[joelshort]

Playing around with ATrousWaveletTransform I think I understand how to make a star mask from that, adding some convolution to the mask and adjusting slightly with curves.  Next up, RangSelection...

[lucchett]

One way consists in applying a strong hdrwt to flatten the stretched image.
This simplify the job for the star mask tool, that is much more effective for small and big stars.
Andrea

[MortenBalling]

@Joel

I have to admit, that after getting started using Harry's excellent videos, I've been using the trial and error method ever since, so I don't know much about available tutorials.

The PI learning curve is so steep, that it's more than vertical in some cases, but hang in there. It gets a lot easier

@Rick

Exactly