#1  
Old 03-11-2013, 03:29 PM
gregbradley's Avatar
gregbradley
Registered User

gregbradley is offline
 
Join Date: Feb 2006
Location: Sydney
Posts: 17,877
Superluminance procedure

Creating a super luminance by adding 1x1 binned RGB to your luminance was discussed at AAIC in July (are those presentations available yet?).

Was that summed? I have tried median combine and don't really see any difference. I did some difference with sum. Not sure if that messes up your LRGB combine though as now you've got sum combined with median. Unless you do all LRGB sum combine. Perhaps that's not a problem.

Anyone using superluminance care to share their procedure?

Greg.
Reply With Quote
  #2  
Old 03-11-2013, 03:40 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
If any use, tried combining red and lum in a recent image when I didn't have much lum - worked well and provided a better result than lum alone on what was a pretty dim red object. Nothing fancy, just put the red and lum files in a single average stack in PI and let it sort out how to normalise them. Regards Ray
http://www.iceinspace.com.au/forum/s...d.php?t=113217

Last edited by Shiraz; 03-11-2013 at 08:53 PM.
Reply With Quote
  #3  
Old 03-11-2013, 09:02 PM
gregbradley's Avatar
gregbradley
Registered User

gregbradley is offline
 
Join Date: Feb 2006
Location: Sydney
Posts: 17,877
Thanks Ray. I've median combined in the past but usually could not notice any improvement. Unless I am doing something wrong. I think it has to be sum combined with Lum to work.

Greg.
Reply With Quote
  #4  
Old 04-11-2013, 08:39 AM
rat156's Avatar
rat156
Registered User

rat156 is offline
 
Join Date: Aug 2005
Location: Melbourne
Posts: 1,694
Hi Greg,

As I wasn't at AAIC, I can't really comment on how to do it, but I might question why you would want to.

I suppose that the theory is that summing the RGB data will give you a "false luminance" image. This is OK in theory, but in practice it makes some assumptions.

Firstly, and you mention it in your question, you have yo shoot your RGB subs at 1x1 binning, which effectively lengthens the time required to acquire the RGB subs vs, say binning 2x2. If you don't do this, then the recombination of that data will reduce some of the hard fought resolution in the L image. So given that you will have to shoot for at least twice as long, probably more to get the same RGB data, why don't you shoot RGB binned and then spend the extra time on L? An example would probably help. Take, for instance, a typical LRGB data acquisition sequence, this may be 4 hours L, followed by 30 minutes of each of RGB binned 2x2, total time taken is 5.5 hours. Now you want to shoot your RGB at 1x1, which will require at least an hour each to achieve the same depth of image, this leaves 2.5 hours for L, the recombination of the RGB data can only add back in an hour of L, so you end up with 30 minutes less luminance exposure.

You would also have to be careful when you took your RGB images, as they now contribute to the resolution of the L image, so all subs now have to be taken whilst the target is high in the sky, this shorten the period you can use to image.

Cheers
Stuart
Reply With Quote
  #5  
Old 04-11-2013, 04:30 PM
alpal's Avatar
alpal
Registered User

alpal is offline
 
Join Date: Jan 2012
Location: Melbourne
Posts: 3,602
All I can say is try it & see if you can improve your pic.

It's more a technique that is handy if you shoot with a One shot colour camera or a DSLR.
In such a case there is more information in the green channel
owing to the Bayer matrix.
The Green channel can be made as a separate channel
& a pseudo luminance layer created.

In Photoshop - go to:
layers
channels
split channels.

cheers
Allan
Reply With Quote
  #6  
Old 04-11-2013, 04:32 PM
gregbradley's Avatar
gregbradley
Registered User

gregbradley is offline
 
Join Date: Feb 2006
Location: Sydney
Posts: 17,877
You bring up some interesting points Stuart.

With regards to 1x1 binning for colour Martin Pugh and Don Goldman and I think R Jay Gabany all said to do colour in 1x1.

I have started doing this on a few images having almost always used 2x2 binning on rgb before (I have done 1x1 on a few occassions).

I think the answer is it depends on what you are shooting and your setup. All these guys are shooting with 3 metre focal length scopes so its a different scenario. With my CDK17 I noticed 1x1 RGB were cleaner, more detailed, less vignetted and cleaned up with callibration better than 2x2 would. So with that setup I think there is a definite improvement over 2x2. As far as how much longer it takes, it does not seem to be 2 or 4 times longer at all. If the data is clean and callibrates better it leads to a nicer image. Martin did recommend longer rgb times than what is typically seen on this site.

Also other factors come into it. Doing a 20 -30 hour masterpiece of every object you intend to image may not be everyone's thing.

Or if you travel to a dark site it most certainly is not a practical approach. Its for those with a home or remote observatory where you can buildup the hours.

But your idea of putting more into the luminance has some merit.

I think 1x1 binned colour makes sense for galaxies or dim objects with fine detail or any object really at long focal length.

But for a widefield perhaps not so much especially at a travelled to dark site where imaging time is precious.

Greg.
Reply With Quote
  #7  
Old 04-11-2013, 08:32 PM
rat156's Avatar
rat156
Registered User

rat156 is offline
 
Join Date: Aug 2005
Location: Melbourne
Posts: 1,694
Quote:
Originally Posted by gregbradley View Post
You bring up some interesting points Stuart.

With regards to 1x1 binning for colour Martin Pugh and Don Goldman and I think R Jay Gabany all said to do colour in 1x1.

I have started doing this on a few images having almost always used 2x2 binning on rgb before (I have done 1x1 on a few occassions).

I think the answer is it depends on what you are shooting and your setup. All these guys are shooting with 3 metre focal length scopes so its a different scenario. With my CDK17 I noticed 1x1 RGB were cleaner, more detailed, less vignetted and cleaned up with callibration better than 2x2 would. So with that setup I think there is a definite improvement over 2x2. As far as how much longer it takes, it does not seem to be 2 or 4 times longer at all. If the data is clean and callibrates better it leads to a nicer image. Martin did recommend longer rgb times than what is typically seen on this site.

Also other factors come into it. Doing a 20 -30 hour masterpiece of every object you intend to image may not be everyone's thing.

Or if you travel to a dark site it most certainly is not a practical approach. Its for those with a home or remote observatory where you can buildup the hours.

But your idea of putting more into the luminance has some merit.

I think 1x1 binned colour makes sense for galaxies or dim objects with fine detail or any object really at long focal length.

But for a widefield perhaps not so much especially at a travelled to dark site where imaging time is precious.

Greg.
Hi Greg,

I think it makes even less sense to not bin images at long focal lengths, oversampling will not positively affect the result. Unless there are very sharp colour changes in your images (other than black to colour around stars), how would seriously oversampling your RGB help the image quality?

Just because some legends do it, doesn't make it right, it just means that you have to argue long and hard to make people change their mind. We also don't know their pixel scale, some of these guys work with exotic cameras that may have huge pixels to match their focal length.

I'd defer to Craig Stark on this one if his opinion differs, he backs his argument up with mathematics.

If, however, you feel better doing it, go right ahead, they are your photos. I have done some tests binning and not binning my RGB, couldn't tell the difference except the not binned images were noisier for the same exposure time, hence I bin my RGB.

Cheers
Stuart
Reply With Quote
  #8  
Old 04-11-2013, 09:07 PM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
Stuart,

Binning relies on the read noise for a binned "super pixel" being a lot less than the read noise for an individual pixel. Sadly, for many of the Kodak/Trusense sensors this is not the case and hardware binning is not that beneficial. You might as well capture the RGB binned x1 and do software binning after the fact if that's what you want.

Cheers,
Rick.
Reply With Quote
  #9  
Old 04-11-2013, 11:02 PM
RobF's Avatar
RobF (Rob)
Mostly harmless...

RobF is offline
 
Join Date: Jul 2008
Location: Brisbane, Australia
Posts: 5,716
I suspect Rick has had lots more experience with this Greg, but in PI at least if you have your unbinned lum extracted from RGB and your original lum, you can integrate them together and PI will proportionally use more of the image with less noise (generally the original lum). I found I got a bit more SNR in the end.

There are posts on the PI forums worth checking out from memory.
Reply With Quote
  #10  
Old 05-11-2013, 09:14 PM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
Quote:
Originally Posted by RobF View Post
I suspect Rick has had lots more experience with this Greg, but in PI at least if you have your unbinned lum extracted from RGB and your original lum, you can integrate them together and PI will proportionally use more of the image with less noise (generally the original lum). I found I got a bit more SNR in the end.

There are posts on the PI forums worth checking out from memory.
Ray mentioned the preferred PI technique above, Rob, but only briefly. If you throw your L, R, G, B masters into ImageIntegration with an average combine and noise evaluation it will weight them based on SNR and integrate them into a Lum+Syn Lum with (hopefully) better SNR than the original Luminance. In my experience it is usually worthwhile. I even did a L+R+G+B+Ha+OIII+SII a little while back and got a really nice low noise super Lum from it.

The nice thing about PI is that you can actually compare the SNR of your results numerically and decide what works and what doesn't even if the difference is subtle visually. Add up a few subtle differences and all of a sudden you can see the improvement
Reply With Quote
  #11  
Old 05-11-2013, 09:56 PM
RobF's Avatar
RobF (Rob)
Mostly harmless...

RobF is offline
 
Join Date: Jul 2008
Location: Brisbane, Australia
Posts: 5,716
Ah - thanks for explaining further Rick. Sounds like a no fuss way to get there, and as you say PI gives you stats if it was worthwhile.
Reply With Quote
  #12  
Old 06-11-2013, 05:38 PM
gregbradley's Avatar
gregbradley
Registered User

gregbradley is offline
 
Join Date: Feb 2006
Location: Sydney
Posts: 17,877
Thanks for the explanations and comments. Its helpful.

So the conclusion is there can be a gain but be careful you don't degrade your luminance in your quest for a bit more SNR. It may be easier to simply get more luminance data and keep it uncorrupted.

Greg.
Reply With Quote
  #13  
Old 06-11-2013, 06:42 PM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
It's "free" so you might as well try it, Greg No harm if you don't use it.
Reply With Quote
  #14  
Old 06-11-2013, 07:52 PM
RobF's Avatar
RobF (Rob)
Mostly harmless...

RobF is offline
 
Join Date: Jul 2008
Location: Brisbane, Australia
Posts: 5,716
And another good reason to image 1x1, possibly getting good RGB data in the bag before going for "perfect lum"? So many ways of skinning the cat in this hobby
Reply With Quote
  #15  
Old 07-11-2013, 04:54 PM
gregbradley's Avatar
gregbradley
Registered User

gregbradley is offline
 
Join Date: Feb 2006
Location: Sydney
Posts: 17,877
Quote:
Originally Posted by RickS View Post
It's "free" so you might as well try it, Greg No harm if you don't use it.
Absolutely. I will have another go. My original question though was what was the best way to combine the luminance and 1x1 RGB's. Sum combine?

Quote:
Originally Posted by RobF View Post
And another good reason to image 1x1, possibly getting good RGB data in the bag before going for "perfect lum"? So many ways of skinning the cat in this hobby
It could be a strategy to always use 1x1 binning with RGB and then chose to bin it 2x2 in Software if you feel its needed and you gain the flexibility to make a super luminance out of the rgb if you want.

Greg.
Reply With Quote
  #16  
Old 07-11-2013, 07:48 PM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
Quote:
Originally Posted by gregbradley View Post
Absolutely. I will have another go. My original question though was what was the best way to combine the luminance and 1x1 RGB's. Sum combine?
An average combination weighted by noise evaluation (as per the PI method I described above) will give you the best results if maximum SNR is what you are looking for. It gives the highest weight to the data with the least noise.

Cheers,
Rick.
Reply With Quote
  #17  
Old 08-11-2013, 05:45 PM
gregbradley's Avatar
gregbradley
Registered User

gregbradley is offline
 
Join Date: Feb 2006
Location: Sydney
Posts: 17,877
Quote:
Originally Posted by RickS View Post
An average combination weighted by noise evaluation (as per the PI method I described above) will give you the best results if maximum SNR is what you are looking for. It gives the highest weight to the data with the least noise.

Cheers,
Rick.
Thanks Rick. That's very good data. Sorry you did post that before.

Cheers,

Greg.
Reply With Quote
  #18  
Old 08-11-2013, 05:47 PM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
No probs, Greg! I used to do complicated PixelMath stuff but the simple ImageIntegration is easier and gives a better result.

Cheers,
Rick.
Reply With Quote
  #19  
Old 08-11-2013, 05:58 PM
gregbradley's Avatar
gregbradley
Registered User

gregbradley is offline
 
Join Date: Feb 2006
Location: Sydney
Posts: 17,877
Quote:
Originally Posted by RickS View Post
No probs, Greg! I used to do complicated PixelMath stuff but the simple ImageIntegration is easier and gives a better result.

Cheers,
Rick.
Mate you are awesome. My unpaid consultant who works out the best practice way of doing something!

Greg.
Reply With Quote
  #20  
Old 08-11-2013, 06:51 PM
Bassnut's Avatar
Bassnut (Fred)
Narrowfield rules!

Bassnut is offline
 
Join Date: Nov 2006
Location: Torquay
Posts: 5,064
Quote:
Originally Posted by RickS View Post
Stuart,

Binning relies on the read noise for a binned "super pixel" being a lot less than the read noise for an individual pixel. Sadly, for many of the Kodak/Trusense sensors this is not the case and hardware binning is not that beneficial. You might as well capture the RGB binned x1 and do software binning after the fact if that's what you want.

Cheers,
Rick.
Really, that suprises me. The whole (only?) point of hardware binning is to reduce readout noise. How do you get to this conclusion?.
Reply With Quote
Reply

Bookmarks

Thread Tools
Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +10. The time is now 03:16 PM.

Powered by vBulletin Version 3.8.7 | Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
Advertisement
Bintel
Advertisement
Testar
Advertisement