How to fix flat-topped (overexposed) stars?

[brew]

Hi,

I am following one fo the excellent tutorials by Gerard (combining HA/Oiii images with LRGB). In this process, I am supposed to take an RGB image and extract the Lab components.

My problem: the extracted a and b components have artifacts in the center of the bright stars, causing subsequent steps to be messed up.

I theorize that the problem is that the RGB bright stars are "maxed out". That is, if I plot them in 3Dplot I see flattened tops to the stars. The rest of the stars are fine. So, I want to do something to fix those bright stars.

I have been trying to generate an image that has a gaussian appearance in the star locations, then use PixelMath to multiply the RGB star data by the gaussian. This not working at all; the gaussians do not have the exact dimensions of the original star, so I get various halos and artifacts. I have tried a number of different masks, both to provide the requisite gaussian shape and to mask the image data in various ways.

I'm thinking my overall approach is wrong. Is there a better way to do this? I was thinking the MorphologicalTransformation might do something, but I don't know what the various parameters do. Is there a simple tutorial (similar to Gerard's PixelMath tutorials) that would explain this tool in general?

Or, perhaps there is a better way to accomplish this to start with?

Thanks for any help,
Robert Brewington

[georg.viehoever]

The general recommendation would be to collect additional shots of this region with shorter exposure times -and thus without burnded out star cores. You can combine these with your existing data using HDRCombination.
Georg

[bitli]

Maybe the script 'Repaired HSV separation' could be useful ? I am not sure it can handle your case.  See http://pixinsight.com/forum/index.php?topic=5149.0. May be you can multiple your image by 0.9 (or something like that) and persuade the script to copy the color information from pixels not saturated to the center part of the star. This is just an idea. It will not make your stars have a nice shape, but maybe a better color.

Regarding the mask, in principle your star is saturated exactly where at least one channel is '1'. So you should be able to build a mask with PixMath with 1 if any channel is saturated.  Maybe grow/fuzzy it a little and try to use on the image multiplied by 0.9 to make the star more gaussian like (some 'deconvolution' or other stuff), avoiding creating more saturated pixels.

-- bitli