Star cores blown out in luminance

[kirkster501]

Hi, My 4 hour stacked luminance has some star and galaxy cores that are slightly blown out even when linear and stretching the image even carefully with HT (Pixinsight) blows them out. I'd like to do a masked stretch but that blows them out of the water completely.

Is there something I can do to stretch that does not blow the star cores please?  Something like DDP in CCD Stack whjere I can keep the cores below 0.85 or so?

Thanks, Steve

[vlaiv]

Why do you worry about luminance star cores?

As for parts of target that gets blown out - use a few shorter exposures. Stack them in separate stack aligned with main stack. Multiply short stack with appropriate value after stacking (ratio of exposure lengths) and "copy/paste" blown out region of the target.

Star cores will certainly reach saturation in luminance after any sort of stretch even if they are not blown in linear data. There is simply no way around it. Dynamic range is just simply too great to be able to see both faint stuff and star cores without blowing out later.

Where people go wrong is color composition. If you compose your image prior to stretch - you will end up with stars that are white in center. Best way to do color composition and retain star color is to:

1. Make sure your RGB subs are not saturated in star cores - use same technique for each channel as described above - take a few short exposures in each channel and mix in with long exposures after stacking. There is simple way to do that via pixel math - "replace long stack pixel value with scaled short stack if over some threshold value - like 90% of max ADU"

2. Stretch your luminance data without color composing until you are satisfied - you can do even things like sharpening / denoising and such at this stage to remove luminance noise.

3. Out of linear, wiped color channels create three images that contain RGB ratio rather than actual color values. First do color calibration. Next stack R, G and B channels via max stacking into temporary image and then divide R, G and B channels with this temporary image. You must be careful not to have negative or zero values in your RGB images - so scale them to 0-1 (or just above 0 to 1) prior to making max stack. After you get ratio images - apply gamma correction (simple gamma 2.2 is enough) on each of them.

4. Produce R, G and B channels of final image by multiplying stretched luminance with each R, G and B ratio images.

Voila, star color in cores will be accurate and not blown out.

[ollypenrice]

Does it have to be in PI? In Ps I'd just do two stretches, one for the bulk of the image and one very soft one for the saturated regions and their immediate surroundings. Then I'd blend them using layer masking.

Alternatively you could experiment with home brewed curves. Essentially you'd bring the curve to a straight line quite low down as below. This would also be the kind of soft stretch I'd use for one of the layers (the one in which we are trying not to further blow the cores.)

I agree with Vlaiv that blown luminance cores are a fact of life but in post processing this can be fixed.

Have you considered using your RGB as a set of short exposures for blending? This often works very well so I often do it for cores. You can convert them to greyscale and treat them as short luminance exposures.

Olly

Edited November 12, 2019 by ollypenrice
Clarification.

[kirkster501]

Ok chaps I will look into that.  Thanks for your thoughts and suggestions.

[wimvb]

CurvesTransformation in PI works the same as curves stretch in PS. Also, have a look here for repairing stars that are blown out:

http://pixinsight.com/tutorials/NGC7023-HDR/index.html#High_Contrast_Small_Scale_Structures