Light Pollution filter setting for Color module

[jhart]

Hi Ivo,

I understand that broadband light pollution filters create data sets that prevent the Color module from rendering good coloring. I live in a very light polluted Bottle 8-9 neighborhood. Using a LP filter such as Optolong L-Pro helps a lot in capturing images that have a considerable amount of that light pollution removed, yet still have detail and most coloring for objects such as galaxies. Would it be possible to have a button in Color or the loading modules that would make StarTools more friendly to LP filters?

Often times I have so much light pollution using no filter or just the IR/UV cut filter, that it seems impossible for Wipe to remove the light pollution yet have a good image for Color and the other modules to work with. I use darks, flats, and darkflats and still cannot approach anything close to a uniform background unless I use one of the Unclibrated buttons. And even then the results are not very good by the time I get to Color. It seems that Wipe will works better on my data captured with the LP filter. However, when I get to the Color module, the colors are way off and I can't find bias slider adjustments to fix them.

Maybe StarTools is too good for me and my so-so data. Just for comparison, Astro Pixel Processor usually does a pretty good job post-processing data captured with a LP filter -- at least on the coloring.

Thanks,
Jeff

[admin]

Hi Jeff,

I think there are a number of misunderstandings here. I hope I can clear some up.

I understand that broadband light pollution filters create data sets that prevent the Color module from rendering good coloring. I live in a very light polluted Bottle 8-9 neighborhood. Using a LP filter such as Optolong L-Pro helps a lot in capturing images that have a considerable amount of that light pollution removed, yet still have detail and most coloring for objects such as galaxies. Would it be possible to have a button in Color or the loading modules that would make StarTools more friendly to LP filters?

Light pollution filters simply reject broad parts of the visual spectrum so that those parts are never recorded. Nor StarTools, nor other software, can "restore" that which was never recorded. As such, it is indeed impossible for the Color module (or any other software's equivalent) to "balance" back signal that does not exist; changing a multiplication bias will not do anything if the the number it is multiplying is 0.

However, StarTools is actually far more friendly to LP-filtered datasets than any other software, by giving you the option to use the luminance signal ("detail") of an LP-filtered stack, while using the coloring of a non-LP-filtered stack (see Compose module).

Often times I have so much light pollution using no filter or just the IR/UV cut filter, that it seems impossible for Wipe to remove the light pollution yet have a good image for Color and the other modules to work with.

This sounds like the perfect candidate for the feature mentioned above. However, it sounds rather strange/impossible(?) to me that - unless you used a mono CCD to acquire the coloring - the coloring would be usable but the luminance would not be, given that they originate from one and the same dataset. If Wipe can clean up the coloring, then the luminance would have been cleaned up equally well (or perhaps I am misunderstanding).

I use darks, flats, and darkflats and still cannot approach anything close to a uniform background unless I use one of the Unclibrated buttons.
And even then the results are not very good by the time I get to Color. It seems that Wipe will works better on my data captured with the LP filter. However, when I get to the Color module, the colors are way off and I can't find bias slider adjustments to fix them.

If the only variable is the LP filter, then the conclusion must be that the LP must be extremely strong and very localized, that you are imaging wide fields, and that the field must be close to the horizon.

The Uncalibrated presets use a high Gradient Aggressiveness to better target the small undulations in a situation where LP is very localized in a wide field, and this is very likely the reason these presets get you closer to what you require.

Maybe StarTools is too good for me and my so-so data. Just for comparison, Astro Pixel Processor usually does a pretty good job post-processing data captured with a LP filter -- at least on the coloring.

Indeed, more basic solutions that use basic sampling setting algorithms as found in APP/Siril/PI, leave it to the user to guess what is light pollution and what is not. In these extreme cases where human nor algorithm can objectively separate celestial signal from terrestrial light, those solutions may achieve "feel good" results, quick. Of course, in reality, real signal will have been removed in the process. The main objection to such simple algorithms is that you cannot be sure how much of the resulting image is the result of wishful thinking, and how much of the image is still true to observed reality. ST's Wipe can, however give you such guarantees; it can guarantee that no signal is removed below a certain undulation frequency, exploiting a 'finite scale' theorem that proposes that celestial detail in your image does not exist beyond a certain scale/size. I.e. anything above that scale is fair game for gradient modelling and removal, while anything below that scale is to be preserved.

Indeed, this very careful treatment of your signal may not come into play so much when dealing with marginal datasets, but definitely becomes very important when dealing with deeper data (e.g. trying to visualise IFN, etc.), where important features/detail would otherwise be compromised by more ham-fisted approaches.

All that said, please feel free to share a stack, and your ST and APP results - perhaps we can give you some pointers with the dataset as-is? In any case, I would highly recommend pursuing the divide-and-conquer approach of acquiring luminance and color separately and see how that works out.

[jhart]

Hi Ivo,

Thank you for that excellent explanation. I will certainly try the recommendation to use a LP filter data set for Luminance in Compose and an unfiltered set for the Color.

Thanks again,
Jeff

[hixx]

Hi Ivo,
I think Jeff and Yourself are looking at 2 sides of the same coin. Sure the really best path is to create a LUM stack using a LPS or even duoband filter plus a Color stack using UV/IR filter. I switched to that method with fantastic results and would encourage Jeff to try so as well.
Having said that, I am living in a Bortle 5 to 6 area not 8 to 9 like Jeff. creating a valid Color stack is challenging for me and might become almost impossible for him when LP is swamping all real data in few seconds.
The only chance might be hundreds of very short exposures for the color stack....
This is why some time ago I came up with the request to provide a "LP hue retuning" feature (now buried in feature requests) which simply shifts recorded energy from green into R and B channels according their correlation when recording thermal objects (stars, galaxies, etc).
The idea behind it is that for those objects, information is still carried in the ratios between the color channels. I. e.: Yellow has a high correlation between red and green (1:1 ratio) hence even if the actual signal energy of the yellows will be gone by the filter, the correlation 8ratio9 won't be. It will be skewed depending on the filter characteristics. To avoid skewing emission objects, You could use that feature on a star mask only or just above a threshold....
In older ST version I abused the Cap Green parameter when it used to cap into yellow (now it seems to cap into white instead). This is as You said just a "feel good" processing, of course. But if some filter's response characteristics would be measured, You could actually use the Matrix to "restore" the yellows by pulling up the R to G and/or G to B ratios to the appropriate values.
I understand its not real science but it might help a lot of users that don't have access to a dark site achieving reasonable "half-gauged" results. As a first step, bringing back the Cap Green into Yellow function might be helpful...
regards,
Jochen

[jhart]

Hi Jochen,

Great post with very interesting ideas. I know imaging from my home in a very light polluted setting is not desirable but I am just trying to make the best of a bad situation. Your suggestion to use shorter color light frame exposures is not something I had considered. I had been pushing longer and longer subs thinking and hoping that would be helpful. I will try shorter ones.

Your color shift filter idea is also very intriguing.

Thanks,
Jeff

[hixx]

Hi Jeff,
Yes, the idea of the short exposures is that the LP gradient should not bring Your sub into saturation. You'll need many of those to limit noise.
You should be fine when the gradient, in the sub's histogram sits somewhere between 30% and 60%. The latter will help recording dimmer objects, of course but requires a stronger Wipe.
Once You have stacked both LUM and RGB stacks against the same reference frame, You want to run these through WIPE individually, as the color stack will require a much stronger WIPE than LUM. Also note You need to employ the exact same crop and BIN methods if used to keep both stacks on top of each other.
Save the wiped results as FITS and reload into LUM resp. R,G,B channels, then process as normal starting with CONTRAST, HDR etc. This should provide decent results, because the remnants of the color stack's LP gradient will be dimmed down by the LUM stack for pixels with no signal.
Good Luck,
Jochen

[jhart]

Hi Jochen,

Thanks for the suggested work flow. However, how do I save the two stacked wiped results as FITS files? The Save function results in TIFF files. After they are loaded in Compose for Lum and RGB, when I get down to the Color module it gives me the warning that Wipe has not been use and renders only a black and white image.

Jeff

[hixx]

I think You can Save as FITS in ST as well by just changing the extension.....
anyone, pls keep me honest....
Jochen

[jhart]

Hi Jochen,

I changed the extension of the two saved TIFF files to FITS. I loaded them in Compose and ST seemed to accept them just fine. However, again when I got to Color the no-Wipe warning popped up and the image was only in black and white.

Jeff

[almcl]

I think You can Save as FITS in ST as well by just changing the extension.....
anyone, pls keep me honest....
Jochen

Afraid not Jochen.

Ivo explained why here:
viewtopic.php?f=5&t=1720&p=7416&hilit=save+fits#p7416

but the relevant bit of the second post is:

"StarTools only saves as TIFF and not as FITS "