Star Saturation - Optimum Exposure for Star Core (Linear)?

JK

Some of my stars are close to saturation and lack colour in the
centre.

 

My camera has a cliff edge drop in noise at Gain 26 which I
like to take advantage of so it looks like I need to reduce exposure (I
currently use 4 or 5 mins for RGB subs).

 

The question is – how saturated should a star be in a linear
sub?

 

I know from Robin Glovers excellent video that I can
theoretically use much shorter subs but with each sub being 120mb, and 12
seconds download/settle time between subs, I want to keep exposures longer than
perhaps is needed.

 

Is there a generally accepted maximum value ( normalised 0
to 0.999) to aim for in the centre of a star in a Linear Sub?

BlockHead

This is a rather lengthy discussion.

I disagree with your assessment concerning stars. It is unreasonable (and unnatural frankly) for star color to be constant in saturation (to the degree that many want). Stars of all colors are naturally desaturated in their centers as represented in digital imagery. You literally need to modify the natural PSF (profile of star light that a telescope or optical system makes) in order to obtain something different. Many image processors do this- but sometimes I feel as an artist it is wholly unwarranted if the object of interest in the field are not the stars. 

The literal answer to your question is, due to the nature of color spaces and how colors are rendered on a screen, the values in star centers need to be below 0.8 (like anything you want to color as I explain in my LRGB videos). But to get to a point you might like... you will end up at a value much less than this. Thus, you will need to kill the star profile to color as you want.

Here is what I just wrote on FB to a similar question:

QUestion: Adam Block what is the suggested stretch here? Regular histogram stretch will blow out the star. Will MaskedStretch work in this situation?

Reply4h

Adam Block

Group expert

Anirban Ray Hmmm... MaskedStretch will give you kinda odd stars (brighter halos with small cores if you have some bright stars in the image). You could blend things.. (but this becomes more advanced/complex). Personally in an example like this, the GALAXY is the important object. I would not process the image to make the stars look good and impact the galaxy. The galaxy comes first in my mind. (I think many people get overly concerned with "blown out"). The way the stars are presented in this image look normal to me. Stars really are bright and "white" in their centers as presented in images. The color is seen more towards the edges of the stars. What I think people are trying to avoid is the harsh representation of the PSF. Normally a star PSF is a smoothly changing (gaussian like) profile from the center to the edges that fade to the darkness of the background. When too much contrast is applied to the stars- they have a harsh edge between the sky and the star which is the look of being "blown out"... so I think avoiding lowering the white point to brighten the image is one thing. Just stretch images with the mid-tones (or curves). Life is a little more challenging with 12-bit or 14-bit CMOS sensors since they have saturated stars quickly- and star saturation kills the nice PSF profile. But does this mean a person should take a billion 10 second exposures to capture NGC 4565 above? NO..that is silly. This will not help the galaxy at all... expose (acquire data) for the best galaxy image... process for the best GALAXY image. That is my suggestion.

-The Blockhead

JK

Thanks Adam, that is useful although I may have given slightly the wrong impression regarding my objective. 

For me, stars are incredibly important and I like to see lots of blue and orange/red stars, but not solid colour across the star, I expect them to have white centres and a smooth progression to a saturated (as opposed to clipped) colour at the edges. 

The problem I am having, and in this case I just realised it is linked to the "NSG losing Blue" problem I posted earlier, is that a lot of stars in my image (crop below) don't have any colour - just white circles. When I examine these they are all at 1.0 and clipped. M101 would look much better if these stars have some colour. I am reasonably happy with the two large stars bottom-right although this was just a quick process.  

The stars that do have colour in the image were originally washed out and thanks to your reply I realised this was due largely to HT.  When I used a small Arc-Sin stretch and then Pixmath 1-(1-$T)
* (1-$T) it kept more colour and allowed a saturation boost to bring out the colour.    

So, I now see two issues, 1) as always, my processing skills, 2) over exposure of stars. 

I will experiment around the 0.8 value you mentioned. If I can get more of the stars below to show some colour I will be happy.