Artifacts in Bright Star Centers

[sreilly]

It's been a while since I have posted in this forum and busy working with data both from home and SRO. The latest project is a deep LRGB image of the Bubble Nebula. I've gotten 42 - 900 second luminance images combined, cropped for the master (dithered), and ABE applied. This gave a very nice and low noise master luminance image but as I proceeded with processing I start to get artifacts. I did a 50 iteration LR deconvolution process that did a good job visually and then proceeded to with a Masked Stretch set at 0.125 and picked a good background area, well as good as you could get considering the star pack field. After applying the stretch I get nice tight stars but the brightest have dark centers. Normally I would protect stars with a star mask for certain processes but I've never heard of doing so for the Masked Stretch process. Am I missing something?

[pfile]

i have seen this on RGB images when one or more channels has saturated star cores - you get these weird pink cores. similar behavior to feeding saturated data to HDRMultiscaleTransform...

did you mask the star cores during deconvolution? it tends to push energy toward the center of the star which makes saturated cores worse; i think they end up more "peaked" than a normal gaussian distribution.

if the star cores are indeed saturated in your original linear image then one strategy would be to take some shorter exposures and use HDRComposition to replace the saturated data.

or i suppose you can try the mask; have never tried that. how many iterations did you do with MS? i have started doing zillions of iterations (even sometimes the max of 1000) to avoid other star artifacts.

rob

[ChoJin]

you might want to take a look at http://pixinsight.com.ar/en/info/processing-examples/28/maskedstretch-stars-sores.html

[sreilly]

Rob,

I think your dead on right and will lower the exposures to 10 minute. I forget how much darker SRO is then here at home in Va. For now I did "repair' the dozen or so stars but don't want to make a habit of having to do so. As far as Masked Stretch, only 1 iteration at 0.125. I also tried 0.16 and 0.19 which did lighten it up a bit more but in the end with the other processes 0.125 seemed to work best for me. Forgive me, I'm thinking 1 instance and that had the default 100 iterations. What improvement are you seeing with ten fold or better?

Steve

[sreilly]

you might want to take a look at http://pixinsight.com.ar/en/info/processing-examples/28/maskedstretch-stars-sores.html

ChoJin,

Thanks for the heads up. I'll look into this repair script, I haven't seen this one before and it looks like it may be helpful if I get the right parameters set.

Steve

[pfile]

Rob,

I think your dead on right and will lower the exposures to 10 minute. I forget how much darker SRO is then here at home in Va. For now I did "repair' the dozen or so stars but don't want to make a habit of having to do so. As far as Masked Stretch, only 1 iteration at 0.125. I also tried 0.16 and 0.19 which did lighten it up a bit more but in the end with the other processes 0.125 seemed to work best for me. Forgive me, I'm thinking 1 instance and that had the default 100 iterations. What improvement are you seeing with ten fold or better?

Steve

in some images i'd get some dark ringing around stars with the defaults, so one day i just pushed it to the max to see what would happen. on a mono image it does not take that long but on a 3-plane image... look out. even on this 4ghz machine it can take 10 minutes to do the stretch to a moderately sized image.

rob

[RickS]

One technique I have used a few times is to do a MaskedStretch, use RangeSelection to build a slightly blurred mask of the saturated star cores and then copy the star cores from a HT stretched image.  You need to match the MS and HT stretches, of course, but that's not hard if you do the HT stretch first then read out the mean value of a background preview with Statistics and use that preview and value as the background target for MS.

Cheers,
Rick.