Help with balancing colors

[da_chaser]

Hello,

I've been experimenting with the free demo version and love the simplicity of this program relative to other options I've tried. I'm able to bring out a ton more detail than I can in Photoshop, but after using Auto Dev I just can't quite get the colors balanced right with a dark, near black background that I'm after. Before I purchase the full version, I want to make sure I'll be able to get the quality images I'm looking for.

Here is a link to a sample stacked FITS file (I personally have been using a TIFF because this file seems to crash my demo version, so I hope it works properly). It's about 90 x 1min subs stacked in DSS with 90 darks plus bias and flats. I use an Astronomik CLS clip-in filter for my Canon DSLR, which I understand can introduce some additional color calibration issues. It's a good sample of my images at this point, with just a couple of months experience under my belt. Definitely a tracking issue probably caused by a poor polar alignment after 90 mins of imaging, and a nasty gradient that is present in all of my images.

https://www.dropbox.com/s/ll2v8uqnq9hxl ... 7.5.14.FTS

Your edited image will be your sales pitch personally to me, so let's see what you got!

[admin]

Hello,

I've been experimenting with the free demo version and love the simplicity of this program relative to other options I've tried. I'm able to bring out a ton more detail than I can in Photoshop, but after using Auto Dev I just can't quite get the colors balanced right with a dark, near black background that I'm after. Before I purchase the full version, I want to make sure I'll be able to get the quality images I'm looking for.

Here is a link to a sample stacked FITS file (I personally have been using a TIFF because this file seems to crash my demo version, so I hope it works properly). It's about 90 x 1min subs stacked in DSS with 90 darks plus bias and flats. I use an Astronomik CLS clip-in filter for my Canon DSLR, which I understand can introduce some additional color calibration issues. It's a good sample of my images at this point, with just a couple of months experience under my belt. Definitely a tracking issue probably caused by a poor polar alignment after 90 mins of imaging, and a nasty gradient that is present in all of my images.

https://www.dropbox.com/s/ll2v8uqnq9hxl ... 7.5.14.FTS

Your edited image will be your sales pitch personally to me, so let's see what you got!

Hi and welcome to the StarTools forum!
What settings did you use in DSS? The data seems to be pre-color balanced (the Red channel ion particular has been pumped up by some extreme amounts), exacerbating noise badly, diminishing StarTools ability to analyse the image effectively and making real color calibration a lot harder. Can you make sure everything is set to 'no' in the stacking steps? Can you confirm you're taking the CR2s from your Canon straight into DSS without any other conversion in between?

Loading of your FTS takes some time as the FTS contains a good amount of 'NaN' (e.g. undefined) values. Such values are interpolated by StarTools (which can take a bit of time). Interpolation of such values helps with avoiding artifacts (due to sudden 'dips' in the signal).

Color calibration is extremely easy in StarTools and much more powerful than any other software (there are many different styles of color composition you can emulate with just a click of a button, while exciting new ones are possible as well). Of course, virgin data is a prerequisite for best results; the better the data, the easier it is to calibrate. StarTools does automatic calibration upon launch of the COlor module, while there are many, many ways to calibrate by hand; by sampling a star field, by sampling a suitable galaxy, by sampling a G2V star, by clicking on a known white light source, or by using the MaxRGB mode to reveal channel dominance (if you don't have a calibrated screen).

If you could get me some data that is more 'virgin', I'd be happy to show you a StarTools-calibrated image!

[da_chaser]

Hi Ivo,

Yes, I take the CR2 files and stack them directly. As far as I can tell none of the processing settings in DSS are enabled, and no processing was done in DSS. Just register, stack, then save as FITS. Could it be caused by my clip-in filter? All of my individual images come out of the camera quite green, but the stacked image does not. So maybe DSS is automatically doing some color balancing on its own?

I also have the autosave file if that helps:
https://www.dropbox.com/s/8dyrq34rd1qdd ... tosave.tif

I wish I could show you how my attempts at processing the image using the demo have turned out. I can get the galaxy to look pretty good but slightly blue, but in the process the background goes red. Too much noise?

[admin]

Hi Ivo,

Yes, I take the CR2 files and stack them directly. As far as I can tell none of the processing settings in DSS are enabled, and no processing was done in DSS. Just register, stack, then save as FITS. Could it be caused by my clip-in filter? All of my individual images come out of the camera quite green, but the stacked image does not. So maybe DSS is automatically doing some color balancing on its own?

DSS is definitely meddling with your data - it's been known to do this (not sure which version or why) even though you can explicitly tell it not to;

Screenshot-Stacking Steps-1.png (36.8 KiB) Viewed 9075 times

What version of DSS are you using?

You could use dcraw from the commandline (see elsewhere on the forum for instructions and correct parameters) to pre-convert the CR2s into 16-bit TIFFs. The 16-bit TIFFs will no longer have the color balancing information attached so that DSS will refrain from making any modifications.

I also have the autosave file if that helps:
https://www.dropbox.com/s/8dyrq34rd1qdd ... tosave.tif

Thanks - it looks the same unfortunately.

I wish I could show you how my attempts at processing the image using the demo have turned out. I can get the galaxy to look pretty good but slightly blue, but in the process the background goes red. Too much noise?

On Windows and (most) Linux desktops, You can use Alt+PrtScr to make a screenshot (can't think of the Mac equivalent right now). Pressing the letter 's' in StarTools should also make a screenshot (undocumented/experimental feature - may or may not work )

[da_chaser]

Well you were right and I found the setting in DSS (RGB channels background calibration was on). Here is a restacked image:
https://www.dropbox.com/s/vw1fvskqu71az ... tosave.FTS

I haven't had a chance to play with it in StarTools yet to see the difference, but I'm curious to see what you can do with it!

[admin]

Thanks!
It's a little easier to process now (still a challenge though!).
This is what I came up with;

M101 Autosave.jpg (270.55 KiB) Viewed 9069 times

The coloring is consistent with the presence of a light pollution filter (a dip in the registration of yellow).
Normally, things to look out for are a good distribution of star temperatures (in a wide enough field). We're missing yellow stars here due the lp filter.
As far as the galaxy goes, you're looking for a yellowish core (older stars and less star formation), bluer outer rims (younger stars and much star formation) in addition to HII areas visible as pink/purple knots. All of these are present, though the yellowness of the core is only very, very slight (again due to lp filter).

With very noisy data like this, color calibration through the usual means (e.g. automation) doesn't work very well as StarTools has too much noise to contend with. However, the field is wide enough for the starfield calibration method (e.g. sample a great number of stars and assume that their average is a good white reference point).
For this, in the Color module, launch the Mask editor and click Auto. Then click the 'Fat Stars' preset. This will select only the brightest of stars (and thus very little, if any, noise). Back in the Color module, click 'Sample' - this will sample only the pixels we have selected (e.g. the stars). You'll notice the red, green and blue sliders changing to point to a new ratio. However, with a mask active, only the selected pixels are affected (e.g. we're now correctly balancing the stars that are selected in the mask) but nothing else. To fix this, go into the Mask editor again and click 'Clear', then 'Invert' - this selects the whole image again. Back in the Color module, the ratios from the previous calibration (e.g. with just the stars) are still intact and so the whole image will now get color balanced.

I bumped up saturation to around 400% to make the color stand out in order to demonstrate its plausibility. No noise reduction was applied yet (this would also affected colors). Hope this helps!

[da_chaser]

Wow, thank you Ivo! I love it and you have sold me!

Any chance I could get you to post your workflow? Also, how would you recommend applying some noise reduction? I'm guessing that would normally be done prior to color correction, which I understand to be the final step. Correct?

Thanks again!

[admin]

Wow, thank you Ivo! I love it and you have sold me!

Any chance I could get you to post your workflow?

Please note that the better the data, the more 'default' the settings become. Right now we're fighting noise and gradients.

--- Auto Develop
Default parameters, to see what we got. We can see oversampling, lots of noise, strong bias.
--- Bin
To convert resolution into noise reduction.
Parameter [Scale] set to [(scale/noise reduction 35.38%)/(798.89%)/(+3.00 bits)]
--- Crop
To frame M101 better (and get rid of stacking artifacts).
Parameter [X1] set to [665 pixels]
Parameter [Y1] set to [496 pixels]
Parameter [X2] set to [1819 pixels (-275)]
Parameter [Y2] set to [1316 pixels (-155)]
--- Wipe
Troublesome gradient that doesn't seem additive in nature; using Vignetting preset (would be good to know what the cause is!).
Parameter [Precision] set to [512 x 512 pixels]
Parameter [Dark Anomaly Filter] set to [4 pixels]
Parameter [Drop Off Point] set to [0 %]
Parameter [Corner Aggressiveness] set to [99 %]
Parameter [Aggressiveness] set to [90 %]
--- Develop
I almost always use AutoDev, but the noise is such that AutoDev has a hard time locking on to 'real' detail. Therefore I'm doing a manual Develop here (I clicked the 'Home In' button repeatedly, but you can obviously choose any sort of setting you like)
Parameter [Digital Development] set to [93.06 %]
Parameter [Dark Anomaly Filter] was set to [5.7 pixels] in order to help Develop set the blackpoint (and ignore dark noise).
--- Color
Procedure as documented above.
Parameter [Saturation Amount] set to [391 %]
Parameter [Blue Bias Reduce] set to [1.26]
Parameter [Green Bias Reduce] set to [1.22]

That's it!

Also, how would you recommend applying some noise reduction? I'm guessing that would normally be done prior to color correction, which I understand to be the final step. Correct?

Noise reduction when Tracking is on (and it *should* be on as long as you can) is performed when you switch Tracking off. You're actually not 'allowed' to do noise reduction until you switch Tracking off, for the simple reason that Tracking, for as long as you can, will yield better noise reduction results. Clicking the Denoise button while Tracking is still on will throw up a popup saying that you're about to get a *preview* of what Denoise can accomplish with the data that Tracking has mined so far, however you won't be able to 'keep' the result just yet (because Tracking isn't over yet and more can be mined).

So, long story short, use Color calibration towards the end of your flow, yes. Then, once you're done, final noise reduction is accomplished by switching Tracking off.

Hope this helps!