Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

Mono camera and luminance filter question

  • Please log in to reply
14 replies to this topic

#1 BlueMoon

BlueMoon

    Soyuz

  • *****
  • topic starter
  • Posts: 3,591
  • Joined: 14 Jun 2007

Posted 23 May 2024 - 09:22 AM

Just mulling over a thought. From my reading the usual path for DSO imaging with a mono camera is LRGB. Generally, an hours worth of luminance data to be combined with red,blue and green filtered data. It appears that luminance lends "detail" to the final image as the RGB data can never be absolutely sharp in focus. Luminance filters pass almost 100% of the visible spectrum but block UV/IR.

 

I'm using this article from wikipedia to provide the base for understanding luminance: https://en.wikipedia.../wiki/Luminance

 

LP is not a concern in this case so no need to filter it out.

 

My question: What if one were imaging a DSO and NOT gathering the RGB data. Only gathering full spectrum data to be processed in grayscale only. Is a luminance filter, outside of blocking UV/IR still useful? Would the monochrome sensor itself capture sufficient detail with the luminance filter simply providing UV/IR blocking? Cheers.


Edited by BlueMoon, 23 May 2024 - 09:31 AM.


#2 SilverLitz

SilverLitz

    Gemini

  • *****
  • Posts: 3,179
  • Joined: 17 Feb 2018
  • Loc: Louisville, KY

Posted 23 May 2024 - 09:37 AM

RGB data is every bit as sharp as Lum, and probably a little sharper as no OTA focuses all wavelengths at the same exact spot (though APOs are close) and the individual R,G,B filters could be focused for the best of each bandpass.  

 

Lum filter's advantage is that there is very little light lost, while each of the R,G,B filters only pass ~1/3 the light, so Lum gathers signal much fast.

 

Even for monochrome imaging, you definitely want the Lum filter's UV/IR blocking, as the focus points for the blocked extremes will be significantly different than the Lum's bandpass.  So, removing the Lum filter will result in less sharpness.


  • BlueMoon likes this

#3 Jared

Jared

    Cosmos

  • *****
  • Moderators
  • Posts: 8,004
  • Joined: 11 Oct 2005
  • Loc: Piedmont, California, U.S.

Posted 23 May 2024 - 09:42 AM

I think your initial statement needs some examination. I would not have suggested that luminance adds “detail” in terms of showing a sharper image than the individual color channels. Rather, using a clear filter as part of the dataset allows one to capture more total photons in a given integration time. More photons allow you to stretch the image more aggressively before noise becomes obtrusive, this allowing you to present fainter structures. This is the additional “detail” provided by capturing full spectrum data through a clear filter. It’s not additional sharpness in the image.

As to whether the clear filter is required if you just want a monochrome image… In most situations, no, it has no benefit. Mostly, it helps keep dust off the sensor chamber window, and it keeps you from having to refocus (or refocus much) vs. the images taken through color filters. If you are using a refractor, it may block some frequencies of light that are not as well focused as most of the visual range (near IR and violet/UV). Not necessary with a reflector, but possibly improves image quality with a refractor and doesn’t hurt anything.
  • BlueMoon and psandelle like this

#4 bobzeq25

bobzeq25

    ISS

  • *****
  • Posts: 34,860
  • Joined: 27 Oct 2014

Posted 23 May 2024 - 09:46 AM

A luminance filter IS a UV/IR filter.  Used to lessen "star bloat" with a scope with lens elements.

 

LRGB is somewhat controversial here (although it's not in my head <smile> ). 

 

My opinion is this.  You don't do L because RGB is not sharp.  You do L to (for a given total imaging time) improve signal to noise ratio.  Using all the pixels on all the light all the time is efficient.  You can get away with less RGB data because of how your eyes work.  The RGB data essentially just "paints" the L.

 

You do pay a small penalty in color accuracy.  L inevitably dilutes color, and the effect is non linear.  So, while you can compensate some in processing, you can't do it perfectly.

 

For some that's a deal breaker, and they shoot straight RGB.  Others (I fall into this group) think the tradeoff is a good one and shoot LRGB.  Particularly in light polluted skies.

 

I do shoot straight RGB on targets with high snr, like globular clusters.

 

Complicated business, this.  Especially color.


Edited by bobzeq25, 23 May 2024 - 09:48 AM.

  • BlueMoon and psandelle like this

#5 BlueMoon

BlueMoon

    Soyuz

  • *****
  • topic starter
  • Posts: 3,591
  • Joined: 14 Jun 2007

Posted 23 May 2024 - 09:49 AM

 

RGB data is every bit as sharp as Lum, and probably a little sharper as no OTA focuses all wavelengths at the same exact spot

That's interesting. I'm thinking then that folks may have been considering chromatic aberration or other imperfections in the glass as the cause for RGB to be "softer" in focus. I suppose then it depends to some extent on how good the optics are.

 

 

Lum filter's advantage is that there is very little light lost, while each of the R,G,B filters only pass ~1/3 the light, so Lum gathers signal much fast.

That's was another consideration over the time it takes to gather RGB data.

 

 

So, removing the Lum filter will result in less sharpness.

Agreed. For the UV/IR filtering certainly worthwhile.  Cheers.



#6 bignerdguy

bignerdguy

    Apollo

  • *****
  • Posts: 1,033
  • Joined: 31 Oct 2019
  • Loc: Lewisville, TX

Posted 23 May 2024 - 09:57 AM

One step i will suggest beyond what the others have said about shooting in LRGB is when you shoot the RGB do it binned as either 2x or 3x.  Then when you go to combine with the Lum resize the RGB frames to match.  What this does is allow you to gather the color data faster and then "paint" the Lum with that to bring out the final details better.  Binning the RGB causes the light to amplify a bit since you are combining each pixel to a newer super pixel (made up of each pixel and some of those around it combined) which while a lower detail resolution has an increased brightness level.   Nice part is that for the RGB frames you don't need to shoot these as for long as you do the LUM frame since the combining to super pixels causes an increase to Signal levels. Not sure I explained it quite right but I imagine you get the idea.


  • BlueMoon likes this

#7 SilverLitz

SilverLitz

    Gemini

  • *****
  • Posts: 3,179
  • Joined: 17 Feb 2018
  • Loc: Louisville, KY

Posted 23 May 2024 - 10:16 AM

That's interesting. I'm thinking then that folks may have been considering chromatic aberration or other imperfections in the glass as the cause for RGB to be "softer" in focus. I suppose then it depends to some extent on how good the optics are.

 

It is the opposite: CA is why Lum would be less sharp than R,G,B (individual filters not OSC).  Lum is getting all the wavelengths with the same physical image plane, and not all of the wavelengths will be at their best focus, while shooting R,G,B individually can have each filter focused for each best focus.


  • psandelle likes this

#8 afd33

afd33

    Viking 1

  • *****
  • Posts: 549
  • Joined: 03 Aug 2023
  • Loc: WI, USA

Posted 23 May 2024 - 10:20 AM

My question: What if one were imaging a DSO and NOT gathering the RGB data. Only gathering full spectrum data to be processed in grayscale only. Is a luminance filter, outside of blocking UV/IR still useful? Would the monochrome sensor itself capture sufficient detail with the luminance filter simply providing UV/IR blocking? Cheers.

You'd end up with a grayscale image with good snr.


  • BlueMoon and bobzeq25 like this

#9 BlueMoon

BlueMoon

    Soyuz

  • *****
  • topic starter
  • Posts: 3,591
  • Joined: 14 Jun 2007

Posted 23 May 2024 - 10:25 AM

 

You'd end up with a grayscale image with good snr.

That is exactly my goal. I want to capture the best amount of detail possible considering the SNR without adding any color information to the mix. This is part of a larger imaging project that I'll be posting about in a month or two.

 

-------------------------------------------------------------------------------------------------------------------------------------------------------

 

So, in a nutshell, based on what you've explained, in my use case where I'm not gathering RGB data, employing a luminance filter really has no particular benefit other than providing UV/IR blocking which is desirable.

 

The mono sensor pixel size, resolution, read noise, amp glow, etc. along with the telescope optics will determine the amount of "detail" in the image. Exposure time, total integration time, seeing, LP presence will figure in as they do for OSC or LRGB imaging I would expect. Cheers.   


Edited by BlueMoon, 23 May 2024 - 10:29 AM.


#10 WadeH237

WadeH237

    Voyager 1

  • *****
  • Moderators
  • Posts: 10,769
  • Joined: 24 Feb 2007
  • Loc: Ellensburg, WA

Posted 23 May 2024 - 10:26 AM

That's interesting. I'm thinking then that folks may have been considering chromatic aberration or other imperfections in the glass as the cause for RGB to be "softer" in focus. I suppose then it depends to some extent on how good the optics are.

Bob is spot on.

 

Here is an example:  Let's say that you have two hours available to image a broadband object with a mono camera.

 

If you are interested in the sharpest possible image, you could spend 40 minutes each on RGB.  The downside to that, is that for dim targets, the data might be quite noisy.  It might be noisy enough that the noise interferes with a sharp appearance.

 

Silverlitz is right above, that RGB is probably a little sharper than luminance.  But in my experience with a good telescope is that the difference isn't detectable (although it might be if you are imaging with an achromat, or even a doublet).  So let's just assume that all the subs are equally well focused and sharp.

 

So instead, you could go for LRGB and spent, say, 1 hour on luminance, and 20 minutes each on RGB.  What this would do for you is to give better S/N in the spatial domain (ie. "sharper") due to more photons passed by the luminance filter.  You could then combine the luminance with the RGB data to get color.  The problem that you might run into is that with only 20 minutes of each color channel, you might have lots of color noise.

 

Back in the day with CCD, you could optionally bin the color channels 2x2 and stick with the LRGB strategy, above.  This would give RGB data with better S/N, but reduced spatial resolution.  This might be a good tradeoff, because it would allow for better color rendition, at the cost of compromising color accuracy at the smallest scales.  To understand why, consider a tiny patch of sky that is 2x2 pixels in the luminance channel.  Each of those pixels might be a different color, but when you bin the color data, those pixels are combined into a single pixel in each color channel.  In the LRGB combined image, those 2x2 pixels cannot therefore end up a different color.

 

In the modern CMOS world, there is no benefit to binning the camera at capture time (except that it makes smaller files, if that is important to you).  Instead, you could just resample the resolution of each of the color channels to match what binning would have done.  Or you could use a different noise reduction strategy on the color channels.  Something like NoiseXTerminator can work wonders on this kind of data.  You could then keep 1x1 color data and mitigate the color noise.  Do keep in mind, though, that any noise reduction will have a smoothing effect that affects data at the smallest scales.

 

There is another option, that I've not thought about too much.  You could capture the color data with 2x2 binning in the camera (even if it's a CMOS camera) and a large dither on every sub.  You could then use CFA Drizzle on the color channels.  That might mitigate the small scale color problem from binning the subs, but I've never really tried this strategy, so I don't have first hand experience with it.

 

Note that this all works due to the way that your biology handles vision.  Your brain gets the spatial details from luminance, so the color data can be low resolution and it will still look good (even if it's not as accurate).

 

For the last couple of years, my own personal strategy has been to do RGB (1x1) and just take many nights of data to handle noise.  This works really well, but lately I am playing with LRGB on many nights of data.  The idea here is to go after tidal streams and IFN in galaxy fields.  I have lots of learn about going after the faintest data, and my processing skills are holding me back...

 

Anyway, I hope that this all makes sense.

 

Edit:  Oh, and to answer your question directly, in the above scenario, you could take luminance for the whole two hours and skip color.  You would have the lowest noise result that your system could manage, but it would be a monochrome image - and there's nothing at all wrong with that.  Monochrome astrophotos can be quite spectacular.  Many years ago, when OSC cameras were not as good as they are today, we often suggested that people start out with a mono camera and just a luminance filter.  We suggested that they could add color filters later, after getting experience with monochrome images.


Edited by WadeH237, 23 May 2024 - 10:30 AM.

  • BlueMoon, psandelle and bobzeq25 like this

#11 BlueMoon

BlueMoon

    Soyuz

  • *****
  • topic starter
  • Posts: 3,591
  • Joined: 14 Jun 2007

Posted 23 May 2024 - 10:37 AM

 

Oh, and to answer your question directly, in the above scenario, you could take luminance for the whole two hours and skip color.  You would have the lowest noise result that your system could manage, but it would be a monochrome image - and there's nothing at all wrong with that.  Monochrome astrophotos can be quite spectacular.  Many years ago, when OSC cameras were not as good as they are today, we often suggested that people start out with a mono camera and just a luminance filter.  We suggested that they could add color filters later, after getting experience with monochrome images.

Yep. I'm revisiting this technique from the past and my intent is to apply some modern colorization techniques programmatically. I'm wanting to ensure that I can get the best results in gathering the raw image data that I can at this point. 

 

I ran across references on the 'net to using a luminance filter but it seemed to apply to LRGB techniques, mostly. One citation mentioned them for mono but I can see now that it was useful in controlling UV/IR and no other benefit. Cheers.


  • bobzeq25 likes this

#12 freestar8n

freestar8n

    MetaGuide

  • *****
  • Freeware Developers
  • Posts: 13,422
  • Joined: 12 Oct 2007
  • Loc: Melbourne, Australia

Posted 23 May 2024 - 06:07 PM

As long as you don't have much chromatic aberration there is no reason to cut out the IR or UV side of the spectrum and you can use a C or clear filter - if you can find one.  I use mainly astrodon filters with EdgeHD11 and have a matching C filter.

 

There are purist arguments that the L band should exactly match the RGB bands for LRGB work - but that makes no sense because the "L" filter is *intended* to correspond to visual luminance - but it doesn't.  Since it is already a poor representative of "luminance" you might as well go all out and suck in all the photons you can and just as accurately call it "luminance."  In both cases - it is just a signal with a bunch of photons in it and does not match what your eye perceives as luminance.

 

I don't think there is much added signal on the blue -> UV side, but there is good signal at 700nm and beyond - usually - particularly for faint and cool stars.  Also for small, faint and red-shifted galaxies.  And there usually isn't much light pollution above 700, so it is fairly clean extra signal.

 

If you try a C filter and find it is noticeably soft with your optics then it's probably not a good idea.  But it works very well for my EdgeHD11 at f/10, so I assume it would work well for others.

 

It's an example of something right there for people to try - but I don't think many do.  You can try out the concept even without the C filter just by leaving an empty filter slot and refocusing as a test.

 

Frank


  • BlueMoon and psandelle like this

#13 BlueMoon

BlueMoon

    Soyuz

  • *****
  • topic starter
  • Posts: 3,591
  • Joined: 14 Jun 2007

Posted 23 May 2024 - 06:58 PM

 

As long as you don't have much chromatic aberration there is no reason to cut out the IR or UV side of the spectrum and you can use a C or clear filter - if you can find one.

Good to know, thanks. I've been reading a bit further with the information you folks have passed along (and some older threads here on CN). If I don't have to deal with UV/IR issues (my Petzval is well corrected as far as I can tell) then I think my initial trials will not include a filter of any sort.  The rationale, from what I understand, is by excluding any unnecessary filtering, I stand the best chance of getting as "deep" as my gear will allow. Cheers.



#14 freestar8n

freestar8n

    MetaGuide

  • *****
  • Freeware Developers
  • Posts: 13,422
  • Joined: 12 Oct 2007
  • Loc: Melbourne, Australia

Posted 23 May 2024 - 07:08 PM

Good to know, thanks. I've been reading a bit further with the information you folks have passed along (and some older threads here on CN). If I don't have to deal with UV/IR issues (my Petzval is well corrected as far as I can tell) then I think my initial trials will not include a filter of any sort.  The rationale, from what I understand, is by excluding any unnecessary filtering, I stand the best chance of getting as "deep" as my gear will allow. Cheers.

I haven't tried it with a refractor and even well-corrected ones may show a halo in blue->UV - so it is certainly possible it may have problems.  But you should find out soon enough.

 

Good luck!

 

Frank


  • BlueMoon likes this

#15 BlueMoon

BlueMoon

    Soyuz

  • *****
  • topic starter
  • Posts: 3,591
  • Joined: 14 Jun 2007

Posted 23 May 2024 - 08:22 PM

 

But you should find out soon enough.Good luck!

Thank you. For me it's not necessarily about success (although that's always nice). I just like to go down different paths to learn and try new stuff. waytogo.gif

 

Cheers. 


  • freestar8n likes this


CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics