ICEINSPACE
Moon Phase
CURRENT MOON
Waning Gibbous 77.4%
|
|
25-09-2013, 07:45 AM
|
|
Space Cadet
|
|
Join Date: Feb 2007
Location: Sydney
Posts: 1,411
|
|
LRGB and CLS light pollution filters test result.
HI Guys.
I asked the question a little while back where it was a good idea to insert a CLS light pollution filter in my imaging train when imaging with LRGB filters as well.
I liver 7.5Km from Sydney CBD so light pollution is an issue for me. I found a site in the UK where someone was doing exactly that, and getting quite good results.
I gave it a go last night, and I have to say I'm pretty happy with it.
I chose the Lagoon because it was a nebula and also because there is a lot of stars and very little black back ground at my focal length. Marc thought that that might be an issue.
The other issue this has solved for me is that I was getting very bloated stars in my blue channel and the CLS filter has cured this issue. I must have be UV light bloating my stars as the CLS has a UV block as well.
The image below has had the following done to it in pix insight only.
R=300secx5 G=300secx5 B=300secx5 No Lum
Stacked with bias and darks only, no flats.
Back ground extraction to remove gradients.
Back ground neutralization.
Colour calibration
histogram
curves.
I'm pretty happy with the result for a quick a dirty test.
http://www.flickr.com/photos/sandy07...8926/lightbox/
Whadaya think??
|
25-09-2013, 08:40 AM
|
|
Registered User
|
|
Join Date: Oct 2011
Location: Canberra
Posts: 951
|
|
Interesting. Shame you didn't do the Trifid, which has a blue reflection nebula and is supposed to be badly affected by the CLS filter. The is a little patch of reflection nebulosity in the lower right of the image next to a brightish star. It's there in your image but definitely knocked back.
You don't seem to have any issues with star colour, but in general everything is more red that we usually see from images of the Lagoon that are quite pinkish.
On the upside your image is lovely.
Personally I'm still conflicted as to whether to get the CLS-CCD or the Hutech LPF-P2.
Thanks for sharing your investigation with us.
Cheers,
CAm
|
25-09-2013, 09:15 AM
|
|
Space Cadet
|
|
Join Date: Feb 2007
Location: Sydney
Posts: 1,411
|
|
Thanks cam, I suspect if I add Lum to the image the redness will wash out a fair bit.
Good point about the trifid if the weather permits I'll take some images of it tonight for comparison. Should have thought about that.
|
25-09-2013, 09:58 AM
|
|
PI popular people's front
|
|
Join Date: Aug 2010
Location: perth australia
Posts: 1,291
|
|
Nice image - however, my complaint with the CLS has always been that you don't ever get yellow stars. Comparing your image to one I took at home without the CLS, the effect is still noticable. I think it's because so much more blue gets through the CLS, yellow stars in the image acquire more blue than they should.
Still - very nice.
|
25-09-2013, 10:42 AM
|
|
ze frogginator
|
|
Join Date: Oct 2007
Location: Sydney
Posts: 22,068
|
|
That's the problem imaging in light polluted areas: blues extinction. Unless you image at the zenith, even with LP filters it's very hard to process back. I'm even having problems with Oiii from my location and that's narrowband. LP filters work to a point but the returns are not that great for the effort IMHO.
|
25-09-2013, 11:36 AM
|
|
Registered User
|
|
Join Date: Apr 2010
Location: Melbourne
Posts: 532
|
|
The thing is that it's physically impossible to restore the narrow band of the spectrum that you've filtered out from the R and G channels - it's irrevocably lost. You can't color balance your way out of it as color balancing means boosting or attenuating whole channels - you might restore the original level of the narrow spectrum that you lost, but you will have boosted the other part of the channel as well. The best you can hope for is a compromise (e.g. some sort of 'off' colour).
The best course of action is to image R, G and B *without* filter and L *with* filter.
Mathematically speaking, light pollution removal is a trivial problem (as long as the data is not clipped because of it) - it's just a bias in the signal that can be removed, so restoring correct colour in R, G and B is not a problem; just model and subtract.
The sole reason why you would want to use an LP filter is to be able to attain longer exposures without the LP swamping the sensor. Since luminance is solely used for the brightness level of your image, not color you get the best of both worlds - it has no effect on colour, while it *has* an effect on how many photons you can gather for your brightness data - best of both worlds!
The only side effect? Objects (ex. stars) that give off light in the spectrum that was filtered appear a tiny bit less bright - they still have the correct colour though, so who cares!
|
25-09-2013, 12:58 PM
|
|
ze frogginator
|
|
Join Date: Oct 2007
Location: Sydney
Posts: 22,068
|
|
Quote:
Originally Posted by irwjager
The best course of action is to image R, G and B *without* filter and L *with* filter.
Mathematically speaking, light pollution removal is a trivial problem (as long as the data is not clipped because of it) - it's just a bias in the signal that can be removed, so restoring correct colour in R, G and B is not a problem; just model and subtract.
|
I'm interested to hear some details and learn a bit more about this. So you say filter the luminance with a LP filter and shoot colour as is. If using an OSC how would you get around that? Also how do you remove the bias in the signal? Is it in the color channels or LUM. With a pixel math operation?
|
25-09-2013, 02:12 PM
|
|
Professional Nerd
|
|
Join Date: Nov 2011
Location: Strathalbyn, SA
Posts: 945
|
|
Quote:
Originally Posted by multiweb
I'm interested to hear some details and learn a bit more about this. So you say filter the luminance with a LP filter and shoot colour as is. If using an OSC how would you get around that? Also how do you remove the bias in the signal? Is it in the color channels or LUM. With a pixel math operation?
|
Using the approach suggested above, if using OSC could you shoot two sets of subs: one set without the filter that is used for RGB, and a second set of subs with the filter on that is then used just to produce the L signal. Then combine the L with the RGB as you would with a mono camera.
|
25-09-2013, 02:21 PM
|
|
Registered User
|
|
Join Date: Apr 2010
Location: Melbourne
Posts: 532
|
|
Quote:
Originally Posted by rmuhlack
Using the approach suggested above, if using OSC could you shoot two sets of subs: one set without the filter that is used for RGB, and a second set of subs with the filter on that is then used just to produce the L signal. Then combine the L with the RGB as you would with a mono camera.
|
Spot on!
To remove bias, yes, pixel math - create a model of the bias (ex. light pollution) and subtract that model from the rest of the image. For the Photoshop afficionados it's a simple matter of subtracting one layer (the bias) from the other (the light polluted image). You must do this while the image is still linear (e.g. unstretched) though, otherwise your colors will still be off.
The trick is of course to model the bias faithfully, but if it's really even across the whole image you could just use a dropper and sample a pixel (or bunch of pixels) that you're absolutely of that they are background pixels (e.g. pixels that should become completely neutral black once the model has been subtracted). Then use the RGB value that you found with the dropper and fill the whole layer with that RGB value. It's quick & dirty but could work reasonably well.
For the more serious work there is ofcourse GradientXTerminator (PS), DBE/ABE (PixInsight), Wipe (StarTools), etc.
|
25-09-2013, 02:33 PM
|
|
ze frogginator
|
|
Join Date: Oct 2007
Location: Sydney
Posts: 22,068
|
|
Quote:
Originally Posted by rmuhlack
Using the approach suggested above, if using OSC could you shoot two sets of subs: one set without the filter that is used for RGB, and a second set of subs with the filter on that is then used just to produce the L signal. Then combine the L with the RGB as you would with a mono camera.
|
Cool. Easy enough. So you debayer the filtered subs direct to LUM and the unfiltered to RGB?
|
25-09-2013, 02:35 PM
|
|
ze frogginator
|
|
Join Date: Oct 2007
Location: Sydney
Posts: 22,068
|
|
Quote:
Originally Posted by irwjager
Spot on!
To remove bias, yes, pixel math - create a model of the bias (ex. light pollution) and subtract that model from the rest of the image.
|
So you debayer filtered to LUM, unfiltered to RGB, extract the LUM and do LUM1 - LUM2 and that's your bias that you can subtract into the color?
|
25-09-2013, 03:20 PM
|
|
Professional Nerd
|
|
Join Date: Nov 2011
Location: Strathalbyn, SA
Posts: 945
|
|
Quote:
Originally Posted by multiweb
Cool. Easy enough. So you debayer the filtered subs direct to LUM and the unfiltered to RGB?
|
That's what I was thinking. Only theoretical for me as i've never used this procedure (not a lot of light pollution out here to worry about) but if i ever move back to the city this is what i'd be trying.
|
25-09-2013, 03:44 PM
|
|
ze frogginator
|
|
Join Date: Oct 2007
Location: Sydney
Posts: 22,068
|
|
Quote:
Originally Posted by rmuhlack
That's what I was thinking. Only theoretical for me as i've never used this procedure (not a lot of light pollution out here to worry about) but if i ever move back to the city this is what i'd be trying.
|
I'll try that next then.
|
25-09-2013, 04:03 PM
|
|
The devil's advocate
|
|
Join Date: Jul 2011
Posts: 816
|
|
Might have to grab one and give it a whirl on my osc, lol i always stick to targets overhead to avoid the issue myself when imaging in the backyard. Would be great to see some side by side shoots from a inner city suburb though.
|
25-09-2013, 05:33 PM
|
|
Registered User
|
|
Join Date: Apr 2010
Location: Melbourne
Posts: 532
|
|
Quote:
Originally Posted by multiweb
So you debayer filtered to LUM, unfiltered to RGB, extract the LUM and do LUM1 - LUM2 and that's your bias that you can subtract into the color?
|
Just to be clear, with an OSC or DSLR;
1. Acquire without filter, debayer as normal, call this your RGB.
2. Acquire with filter, debayer as normal, call this your L
3. Convert L to grayscale (*average* the channels, don't use lightness, luminance, etc. - you'll want to weigh each channel the same, i.e. 33.3%), now you got your L, ready to go.
4. Create a model of the light pollution that exists in your unfiltered RGB image (keep your RGB linear!), call this M.
5. Put RGB in a layer, put M on top, set M's layer to 'subtract'.
6. Save the 'RGB-M' result as your new RGB; RGB without light pollution.
7. Process and combine L and light pollution-free RGB with your favorite method.
Happy to give anyone a hand if you got the data!
|
25-09-2013, 06:22 PM
|
|
Registered User
|
|
Join Date: Jan 2011
Location: Rylstone, NSW, Australia
Posts: 1,426
|
|
Ivo,
Much appreciate what you are saying here. I'm using a mono CCD with the LP filter only for luminance. However, I find my LP is in quite a gradient and certainly not smooth over the whole image. How would you process the RGB stacked frames to eliminate the gradient? I've taken note of the requirement to do this before stretching (but without stretching I find it somewhat difficult to see the gradient). My usual tools are CCDStack and CS-5. I've had some success using the flatten background tool in CCDStack before creating the RGB combine in CCDStack. However, I don't always have success with that tool. Thanks for any assistance!
Peter
|
25-09-2013, 11:17 PM
|
|
Space Cadet
|
|
Join Date: Feb 2007
Location: Sydney
Posts: 1,411
|
|
Interesting discussion. I guess this technique could be translated into LRGB imaging with a mono ccd. The principles should be the same, it will just need a bit of thinking through.
Thanks for the input.
|
26-09-2013, 07:25 AM
|
|
Registered User
|
|
Join Date: Apr 2010
Location: Melbourne
Posts: 532
|
|
Quote:
Originally Posted by PRejto
I find my LP is in quite a gradient and certainly not smooth over the whole image. How would you process the RGB stacked frames to eliminate the gradient? I've taken note of the requirement to do this before stretching (but without stretching I find it somewhat difficult to see the gradient). My usual tools are CCDStack and CS-5. I've had some success using the flatten background tool in CCDStack before creating the RGB combine in CCDStack. However, I don't always have success with that tool.
|
Honestly, my knowledge of the latest versions of PS is really quite limited, but I'm assuming you can 'layer' a stretch on top of a linear image layer. I imagine you should be able to 'sandwich' your model subtraction in between, so that you can see what you're doing, even though all the data is still linear - the 'stretch' that is layered on top will help you see. When you're ready to finalise the light pollution-less image, you simply remove the stretch from the layer stack and flatten the image.
I've heard FITSWork (free) has a gradient removal/modeling tool that you could try. And GradientXTerminator is a Photoshop plug-in (paid) for this very purpose.
In the meantime, if you've gathered the data and would like a hand with the modeling, let me know. I'm pretty keen to try this technique myself with some data - one of my users reported success with technique when I suggested it, so I'm pretty keen to replicate it!
|
26-09-2013, 08:17 AM
|
|
ze frogginator
|
|
Join Date: Oct 2007
Location: Sydney
Posts: 22,068
|
|
Quote:
Originally Posted by irwjager
4. Create a model of the light pollution that exists in your unfiltered RGB image (keep your RGB linear!), call this M.
|
That's the only step I'm not clear with. How do you do that?
|
26-09-2013, 10:33 AM
|
|
Registered User
|
|
Join Date: Apr 2010
Location: Melbourne
Posts: 532
|
|
Quote:
Originally Posted by multiweb
That's the only step I'm not clear with. How do you do that?
|
That's the trick of course, but if it's really even across the whole image you could just use a dropper and sample a pixel (or bunch of pixels) that you're absolutely certain of that they represent background pixels (e.g. pixels that should become completely neutral black once the model has been subtracted). Then use the RGB value that you found with the dropper and fill the whole layer with that RGB value. It's quick & dirty but could work reasonably well.
For the more serious work there is ofcourse GradientXTerminator (PS), DBE/ABE (PixInsight), Wipe (StarTools), etc. that either remove light pollution or can create a model for you.
|
Posting Rules
|
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
HTML code is Off
|
|
|
All times are GMT +10. The time is now 02:07 PM.
|
|