Back to forums

[RCC] NGC3718 - more „Pop“, better stars?

LRGB Request for feedback NGC3718 NGC Objects Deep Sky
5 replies•232 views
Torben van Hees avatar
Torben van Hees avatar

NGC3718

https://astrob.in/oiujzm/0/

Hello, I have combined data from last year (about 6 hours) and this year (8 hours) LRGB (3-1-1-1) for this target. All four days that I collected the data had subpar seeing (3“), so I know the amount of detail possible is limited. It‘s from Bortle 5 and 6. I will try to gather more data but weather is looking bleak again.

Workflow was:
WBPP
MureDenoise on all masters (0.6)
stack all masters for superlum (really improved on the SNR)
Crop borders, realign to superlum
Superlum:
Deconvolution
TGVDenoise
MMT with Denoise
MaskedStretch w/o clipping
ArcSinH Stretch with star mask

Channel combination of RGB.
Photometric color calibration
MLT denoise
ArcSinH Stretch
ACDNR

LRGB combination
HDRMultiscaleTransform
HT
Curves (Lum, Saturation)

I am not satisfied because of two reasons:
One is the colors, I think there should be far more blue hues, but I just cannot get it to pop out more in this image no matter what I do in curves. Maybe too little RGB data, but I do not feel so had this problem before as severely.

The other is the stars: They bloat during stretching and I have not found a good workflow for this image to prevent that. I got them to this point with masking and a selective curves transformation but I do not like them still. One possibility is, of course, that they are too bloated to fix because of seeing.

Any tips?
Helpful Engaging
Rod Kennedy avatar
Rod Kennedy avatar
Hi Torben

It is tempting when stretching stars to stretch the peak close to 1.0 (to utilise the full dynamic range) but this washes out the color of bright stars.  In fact your stars have good color information.  To see this (just as a demo): in CurvesTransformation set RGB/K linear from (0,0) to (0, 0.8 ) so that the maximum luminosity now becomes 0.8 instead of 1.0 (remember in PixInsight the image is floating point so you aren't throwing away information). Or easier use PixelMath with 0.8 * $T. Simultaneously strongly boost the saturation S channel.  Then you can see there is color information.

I inferred this stuff by inspecting Astrobin images with Christmas tree light stars.

Rod

Helpful Insightful
Torben van Hees avatar
Torben van Hees avatar
Thank you for that tip concerning star color. I did not think about that yet but it is very logical. As you approach 1.0 the differences between the colors will diminish even if you are not clipping. I think ArcSinH stretch will partly prevent that and have had good results (or so I like to think) in the past. I will try not to saturate them too much. Combined with another tip I got elsewhere that might fix the washed-out colors.
Respectful
Byron Miller avatar
Byron Miller avatar
Is the masked stretch messing with stars?  arcsinh should stretch star color beautifully but I've never done it with a masked stretch prior.
Torben van Hees avatar
Torben van Hees avatar
I‘m not sure. After the masked stretched the stars look „ok“ - I really fear the data is not up to snuff. I am not much concerned about star color - I can work that out. I think the galaxy is „drowning in browns“. The ArcSinH stretch bloats the stars, but the masked stretch brings out too much noise so I tried to combine both. This was the best compromise I found. Don‘t like it much.
Byron Miller avatar
Byron Miller avatar
To minimize star bloat, instead of doing one large arcsinh stretch, I'll do a decent size one, then smaller and then tiny - that  seems to work better than one large stretch for me