How do you know that your image is well color balanced?

[topboxman]

I always run DBE followed by BN and CC. How do I know that my image is well color balanced? Do you use STF's link/unlink color channels to make a judgement by blinking in between link and unlink color channels and notice no noticable change in color?

I guess another way is to stretch with Histogram until you see all three R, G, & B channels line up but I am not sure that's accurate.

Also, when you apply STF to Histogram, do you use STF's unlink or link color channels? I would assume STF's unlink color channel because it looks more correct than STF's link color channels.

Thanks,
Peter

[Geoff]

I always run DBE followed by BN and CC. How do I know that my image is well color balanced? Do you use STF's link/unlink color channels to make a judgement by blinking in between link and unlink color channels and notice no noticable change in color?

I do this as a bit of a check, but usually assume that linked channels is giving the desired result

I guess another way is to stretch with Histogram until you see all three R, G, & B channels line up but I am not sure that's accurate.

Another thing I check up on.  Almost always the RGB channels line up after colour calibration.  If they don't I may (or may not) give a tweak to the histogram.

Also, when you apply STF to Histogram, do you use STF's unlink or link color channels? I would assume STF's unlink color channel because it looks more correct than STF's link color channels.

I apply linked colour channels.   After all, BN and colour balance are supposed to render the linked stretch as the "correct" one (whatever that means).
Geoff

[topboxman]

When I said "correct" I meant by visual inspection that the colors appear to look more correct with STF's unlinked channels than linked.

In my experience, the colors in STF's unlinked channels usually looks more correct (or better balanced) than linked whether the images have been color calibrated or not. If the color calibration is perfect (or near perfect), I can tell by comparing STF's linked and unlinked color channels and there's no change in between the two, then I assume the color calibration is done right.

Sometimes after DBE, BN and CC and I cannot get STF's linked/unlinked color channels to match, then I use STF's unlinked channels and apply to Histogram Transformation. It could possibly be that some of my images may have low SNR and may be more difficult to balance the colors.

If the auto-stretch in STF is aggressive and shows lots of noise, is it possible to reduce the aggressiveness of STF's auto-stretch so I can apply less noisy looking image to HT?

I always image from fairly high light pollution area.

Thanks,
Peter

[Geoff]

When I said "correct" I meant by visual inspection that the colors appear to look more correct with STF's unlinked channels than linked.

Sorry for causing a misunderstanding Peter. I wasn't actually referring to your use of the word.  It was meant as a reference to the type of discussion about the reality of colour in CCD images that went on in this thread http://pixinsight.com/forum/index.php?topic=2542.0

If the auto-stretch in STF is aggressive and shows lots of noise, is it possible to reduce the aggressiveness of STF's auto-stretch so I can apply less noisy looking image to HT?

Yes you can, but you will probably have to zoom into the STF window to spread the sliders.  Click the magnifying glass with the + sign, then click in the window.  Then click on the arrow sign.  You can then move the sliders around to adjust the stretch.
Geoff

[papaf]

You can also CTRL+click on the nuclear icon to set the desired median values for background and highlights.

[jkmorse]

When I said "correct" I meant by visual inspection that the colors appear to look more correct with STF's unlinked channels than linked.

In my experience, the colors in STF's unlinked channels usually looks more correct (or better balanced) than linked whether the images have been color calibrated or not. If the color calibration is perfect (or near perfect), I can tell by comparing STF's linked and unlinked color channels and there's no change in between the two, then I assume the color calibration is done right.

Peter, can you (or anybody else that wants to jump in) talk a bit more about the impact/theory of linking or unlinking channels in STF?

Thanks,

Jim

[topboxman]

Hi Jim,

Good question but I don't think I can answer which is why I created this thread. 

Peter

[pfile]

unlinked STF simply brings the center value of each channel's histogram to the same place. so the channels 'perfectly' overlap and i guess the color seems the most neutral/pleasing.

that does not necessarily mean the features that should be considered white have been made white, but overall the image is 'white' if that makes any sense.

rob

[jkmorse]

Thanks, that helps.  I always find that I need to tweak the colors after my HT stretch to get the black point right and this could be the cause.

I would be interested in seeing how this works with the whole color calibration routine and whether the experts recommend one setting over the other.

Jim

[pfile]

after you do CC you should only apply STF with the channels locked, or else it will undo all the hard work you've done

rob

[topboxman]

after you do CC you should only apply STF with the channels locked, or else it will undo all the hard work you've done

rob

Even if STF's unlocked channels may look better?

Peter

[Harry page]

Hi
Here is I see it in most cases

when RGB are combined it is very rare that there is a pleasing colour balance , so we unlink the channels to get a better look at what the data might contain
after this I Run DBE and when done well a neural background is achieved and usually a reasonable colour balance is achieved.
At this stage the channels must be linked on the STF to give a correct view of the Data "to date" if further colour balance is required I then run colour calibration
( still with linked channels on the STF) , which in 99% of cases is all that is required   
The key here is getting DBE right , sometimes a increase in tolerance will remove any colour imbalance  and when correct there will be no difference between having the stf channels linked or unlinked .
However in the 1% of cases where a bias remains , (usually caused by a very weak channel and seen a lot in DSLR Data) a better image seems to remain with the STF channels unlinked so to correct this ( while linear , usually after DBE)
1) split the RGB into separate channels
2) Select the linear fit tool
3) Select a strong channel as a reference
4) Apply to the other 2 channels
5) recombine 
And The last 1% of images are sorted 

The background of a image should always be neutral and arbitrary adjustment of colours with curves is to be avoided IMO

Regards

Harry

[pfile]

Even if STF's unlocked channels may look better?

yes. what CC is doing is figuring out the relative strengths between the channels in order that features that should be white are white. in the case of a galaxy, those features are all of the stars (typically unresolvable) in your preview. in the case of some other DSO image (structure detection turned on in CC) it's all the stars in the image (stars which are inside our galaxy).

in order for this to work right, the background has to first be neutralized. this sets the stage for CC by removing any gross color casts in the image. DBE generally leaves the image with a near-neutral background (but only when "normalize" is turned off, as is the default.) still, i always follow DBE by BackgroundNeutralization, just to be complete.

so if after CC you do something to the image to change the channel median values (change individual channels in HT, apply an uncoupled STF to HT), well, you changed the color balance of the image, making it "wrong" again.

WRT color balancing a galaxy, you need to be careful to make the preview big enough to capture a wide range of stars. the stars in the center of a galaxy are typically older and redder than young stars (bluer, hotter) in the arms. if you only put your preview on the very center of the galaxy you're going to make those reddish/yellowish stars white, which is not really correct.

i see harry's reply now and that's a good point; a weak channel (with lots of noise in the background) can mess things up (BN picking up pixels which are just noise). and as harry says, after DBE typically linked/unlinked STF shows less of a difference due to the removal of the casts.

rob

[topboxman]

Thanks Harry. Your explanation is very clear and makes sense.

Also, thanks Rob for a detailed explanation as well.

Peter