#1  
Old 29-05-2016, 04:40 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
what should stars look like?

Stars are points of light with no size when viewed at interstellar distances and amateur apertures. Therefore, everything that results in round stars of varying size is an aberration. The main contributor to star size (and variation) is the seeing/guiding, which causes the points of light to wobble around and spread out into (not quite) 2D Gaussian structures. Bright stars look big because the stretching brings up more of the skirts of the bright star profiles. The relative "bigness" of the stars depends on the size of the star profile (atmospheric), the level of stretch and the sampling - fine sampling gives big stars.

On that basis, what should we aim for? I have been assuming that, when an image is stretched to bring out a dim target, the bright top end of the scene will be correspondingly compressed and brighter stars (whatever colour) will turn into almost fully saturated white disks with small skirts of unsaturated colour - as happened with film. However, I notice that many here use various forms of masked stretch to preserve some of the original profile and colour of the stars, while stretching the rest of the image (ie the image becomes a composite - one image for the stars and one for the rest). In the process, the stars taken on a fuzzier, softer appearance, with more colour.

My philosophy for now is to prefer relatively sharp edged stars (ie no masking during stretching) because they are much easier to distinguish from background galaxies, but there is clearly no right answer and I would really appreciate the opinions of others.

Last edited by Shiraz; 29-05-2016 at 07:48 PM.
Reply With Quote
  #2  
Old 29-05-2016, 05:01 PM
Slawomir's Avatar
Slawomir (Suavi)
Registered User

Slawomir is offline
 
Join Date: Sep 2014
Location: North Queensland
Posts: 3,240
As a priority, I always aim for the best data I can get with my gear (round crisp stars in the subs are my best indicator), then I honesty focus entirely on the nebulosities and fuzzies- at this stage in my career as an amateur astrophotographer I must admit I completely neglect processing stars and try to apply any tools very gently to preserve the original data as much as possible, that of course does not include stretching and colours, both of which I often overdo
Reply With Quote
  #3  
Old 29-05-2016, 08:54 PM
billdan's Avatar
billdan (Bill)
Registered User

billdan is offline
 
Join Date: Mar 2012
Location: Narangba, SE QLD
Posts: 1,551
What I have been doing ( rightly or wrongly), is global stretch first and then I protect the stars with an inverse mask for modules like Decon, wavelet sharpening and HDR etc.
I keep the mask on until I do the colour balance and saturation.

After that I may put a star mask back on to either shrink the stars or fix up any issues with roundness.

Bill
Reply With Quote
  #4  
Old 29-05-2016, 09:30 PM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
A little soft and colourful is my preference but I don't mask the stars. An Airy disk ain't hard!

Cheers,
Rick.
Reply With Quote
  #5  
Old 29-05-2016, 10:24 PM
Atmos's Avatar
Atmos (Colin)
Ultimate Noob

Atmos is offline
 
Join Date: Aug 2011
Location: Melbourne
Posts: 6,980
I have not yet played with a Masked Stretch but I do tend to largely process the stars and everything else separately, mostly because if I don't mask the stars I badly bugger them up

I personally prefer the softer approach to stars most of the time, I just haven't really gotten close to mastering it.
Reply With Quote
  #6  
Old 30-05-2016, 02:38 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
thanks guys - comforting that there are many different opinions.

Rick, doubt we ever get close to dealing with Airy disks - wish the seeing was that good - even occasionally would do.

my basic problem I guess is that "really bright" on the screen implies that all pixels of all colours are flat out - surely that means that really right stars cannot have colour in an image that tries to use the whole dynamic range? A pure colour can only have about 1/3 max brightnesss?
Reply With Quote
  #7  
Old 30-05-2016, 04:01 PM
janoskiss's Avatar
janoskiss (Steve H)
Registered User

janoskiss is offline
 
Join Date: Jan 2005
Location: Sale, VIC
Posts: 6,033
With a perfect optics and detector stars should look like Airy disks. If you use a telescope other than a refractor, they should look like the equivalent of an Airy disk for your scope (in more precise technical terms: the square magnitude of the Fourier transform of the aperture function of your scope).

But with practical detectors (CCDs etc) the signal will bleed to neighbouring pixels. That's both good and bad: good because in the image brighter stars will look brighter; bad because the extreme brightness of stars is only captured by bloating their imprint on the pixel array. It's a dynamic range issue and can to some degree be tamed with HDR techniques.
Reply With Quote
  #8  
Old 31-05-2016, 06:55 AM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
Quote:
Originally Posted by Shiraz View Post
thanks guys - comforting that there are many different opinions.
Yes, we make a lot of aesthetic decisions on how we portray the data we capture and this is just another one, IMHO

Quote:
Originally Posted by Shiraz View Post
Rick, doubt we ever get close to dealing with Airy disks - wish the seeing was that good - even occasionally would do.
Get a FSQ-106, Ray! Your Airy disk in red would be about 3 arc sec.

Or, move your gear to SRO Our Ceravolo 300 at SRO should have an Airy disk a little over 1 arc sec and the seeing is often better than that.

Quote:
Originally Posted by Shiraz View Post
my basic problem I guess is that "really bright" on the screen implies that all pixels of all colours are flat out - surely that means that really right stars cannot have colour in an image that tries to use the whole dynamic range? A pure colour can only have about 1/3 max brightnesss?
I'm happy with white cores on bright stars. I just like to see some colour around the core. As David Malin tells us, stars are black body radiators and should have pastel colours, so they can be brighter than 1/3 max.

Cheers,
Rick.
Reply With Quote
  #9  
Old 01-06-2016, 09:42 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by janoskiss View Post
With a perfect optics and detector stars should look like Airy disks. If you use a telescope other than a refractor, they should look like the equivalent of an Airy disk for your scope (in more precise technical terms: the square magnitude of the Fourier transform of the aperture function of your scope).

But with practical detectors (CCDs etc) the signal will bleed to neighbouring pixels. That's both good and bad: good because in the image brighter stars will look brighter; bad because the extreme brightness of stars is only captured by bloating their imprint on the pixel array. It's a dynamic range issue and can to some degree be tamed with HDR techniques.
Quote:
Originally Posted by RickS View Post
Get a FSQ-106, Ray! Your Airy disk in red would be about 3 arc sec.

Or, move your gear to SRO Our Ceravolo 300 at SRO should have an Airy disk a little over 1 arc sec and the seeing is often better than that.

Cheers,
Rick.
OK, agree that there is a diffraction contribution, but I still think that atmospheric seeing overwhelmingly determines the star profiles most of the time.

just resurrected an old Excel model to check (it isn't perfect, but I tested it against the measured PSF of my old 200mm scope and think it works well enough - would gladly be corrected if anyone has better). The attached image shows the star profiles from three scopes (FSQ106, 250mmNewtonian with 0.3 central obstruction and 400mm with 0.35 CO). Three atmospheric seeing conditions were considered (0.1arcsec - perfect, 1arcsec - exceptional and 3arcsec - ordinary). The model is the standard 2xBessel with convolution of a Gaussian to represent seeing (should really have used Moffat, but Gaussian is close enough for this task). Note that the y-axis has a log scale.

Even in 1 arcsec seeing, you don't see an Airy disk structure and the scopes all have significant regions where the star shape is primarily determined by the seeing and not by diffraction. In 3 arcsec, diffraction does not get a look-in for determining the shapes of most stars - it's all down to seeing for 3-4 orders of magnitude from the peak (ie after stretching, all but the brightest stars will have profiles dominated by seeing). The seeing-dominated part of the profiles is quite steep, so I would expect stars in that brightness zone to have fairly tight boundaries

All of these scopes produce star profiles with skirts (due to diffraction) and there will also be some scattering (not considered). The larger scopes will give dimmer skirts around the bright stars, but with all, the increase in star size with bright stars or long exposures is due to this skirt structure and not any charge overflow or blooming in the sensor (most sensors have potential barriers and overflow drains specifically to prevent this from happening).

The skirts are brighter on the FSQ - this accounts for the big cotton balls these scopes produce around very bright stars. This is not a fault and can look very attractive - it is just what happens with smaller apertures. The larger scopes produce dimmer skirts, so stars from these will only grow very large when the exposure is long enough and the stretching is sufficient to bring the dim skirts into the scene dynamic range.

well anyway, that's how I see it and would welcome any corrections/comment. regards Ray
Attached Thumbnails
Click for full-size image (scopes.jpg)
168.5 KB47 views

Last edited by Shiraz; 01-06-2016 at 11:38 PM.
Reply With Quote
  #10  
Old 02-06-2016, 01:19 PM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
Very interesting result, Ray. I've often wondered what contributes to star bloat and halos.

What effect does the brightness of the star have? Does it have the same proportional effect on the centre and the skirts?

Have you thought about applying a non-linear stretch, say a PI style MTF stretch, to your results and then plotting a linear result?

BTW, those stars look pretty soft to me... though the log scale helps

Cheers,
Rick.
Reply With Quote
  #11  
Old 02-06-2016, 02:20 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by RickS View Post
Very interesting result, Ray. I've often wondered what contributes to star bloat and halos.

What effect does the brightness of the star have? Does it have the same proportional effect on the centre and the skirts?

Have you thought about applying a non-linear stretch, say a PI style MTF stretch, to your results and then plotting a linear result?

BTW, those stars look pretty soft to me... though the log scale helps

Cheers,
Rick.
These are normalised to the peak, so the same profile will apply to any star, but the absolute value will vary.

When you stretch an image to reveal, say, a galaxy, you may be expanding a region of perhaps 2 orders of magnitude so that it covers the whole dynamic range of the image. If you have moderately bright star, that 100:1 band may now be aligned somewhere around the middle of the the star profile, where there is precipitous fall and the star will vary very little over the 100:1 intensity range of interest. Thus, the star edge will be fairly sharp and the core will be saturated - even though the core did not start out saturated. If the star is dim, the band of interest will align with the top end of the profile and you will see a distorted Gaussian profile in the core and a fairly sharp edge. If the star is very bright, the band of interest will be down the bottom of the star profile and the star will have a large soft skirt and a large saturated core (again, even though the core did not start out saturated).

Doing a PI-type stretch is a great idea - much better than the words above. Now to figure out what types of functions are used for stretching - any ideas?

regards Ray

Last edited by Shiraz; 02-06-2016 at 03:38 PM.
Reply With Quote
  #12  
Old 02-06-2016, 02:30 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 Shiraz View Post
These are normailsed to the peak, so the same profile will apply to any star, but the absolute value will vary.
Thanks, Ray. That's what I suspected.

Quote:
Originally Posted by Shiraz View Post
Doing a PI-type stretch is a great idea - much better than the words above. Now to figure out what types of functions are used for stretching - any ideas?
The doc for HistogramTransformation (yes, there is some!) explains the MTF in gory detail. That would be a good place to start, IMO.

Cheers,
Rick.
Reply With Quote
  #13  
Old 02-06-2016, 05:17 PM
Slawomir's Avatar
Slawomir (Suavi)
Registered User

Slawomir is offline
 
Join Date: Sep 2014
Location: North Queensland
Posts: 3,240
Really interesting results that make me think my next scope could be a fast Newtonian. Thank you Ray for taking time to do the hard work for us :-)

As for functions used in stretching, I would also need to read the info on HistogramTransformation; however, apart from being hopelessly dependent on PI, I also like experimenting with free FITS Liberator since it allows for a few different functions - please refer to the attached print screen.

Cheers,
Suavi
Attached Thumbnails
Click for full-size image (FITS.jpg)
127.7 KB46 views
Reply With Quote
  #14  
Old 02-06-2016, 05:26 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by Slawomir View Post
Really interesting results that make me think my next scope could be a fast Newtonian. Thank you Ray for taking time to do the hard work for us :-)

As for functions used in stretching, I would also need to read the info on HistogramTransformation; however, apart from being hopelessly dependent on PI, I also like experimenting with free FITS Liberator since it allows for a few different functions - please refer to the attached print screen.

Cheers,
Suavi
just what I need. Thanks very much Suavi
Reply With Quote
  #15  
Old 03-06-2016, 07:47 AM
gregbradley's Avatar
gregbradley
Registered User

gregbradley is offline
 
Join Date: Feb 2006
Location: Sydney
Posts: 17,877
I think you also need to consider the well depth of the CCD/CMOS detector in this equation. Also how well the antiblooming works. Antiblooming is supposed to prevent bleed from too much exposure going over into the surrounding pixels.In practice I doubt it works that perfectly and not at all binning levels. For one thing its in the hands of the camera manufacturer to set the antiblooming level and for another electronically I doubt anything works perfectly 100% efficiently. It would also be hidden thing about how well the antiblooming is implemented on the circuitry of the sensor itself.

I have noticed very clearly time and again that cameras with deep wells protect the stars way better than cameras with small wells. Stars get wrecked 5X faster with small welled cameras than deep welled cameras.

The KAF16803 has around 100,000 electron full well depth. Its hard to wreck the stars with that camera. The trend is to CCDs and CMOS with small pixels and thus small wells. Users of those cameras need to be more careful about stars and stretching data etc.

You see this mainly with DSLRs where its common to see a DSLR image that hasn't taken this into account when processing and all the stars are white with no colour, ie. all overexposed.

Greg.
Reply With Quote
  #16  
Old 03-06-2016, 09:55 AM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
I can't see any reason why well depth is an issue at all. I think that, provided you don't overexpose by using excessive sub length, the star profiles will be exactly the same for pixels with small wells or large wells. They will look bigger if the pixels are small, because there will be more pixels under the star profile, but the shape will be the same.

Well depth is a factor in star shape if you use subs that are too long - and then you will saturate the cores and bring up the dim skirts. This is easy to do with chips with small wells if you try to use long subs (eg with your 694 and typical optics, anything much over 5 minutes in luminance will probably be too long) - but you don't need to use long subs if the read noise is low and the dynamic range will still be good with short subs. If stars are being wrecked 5x faster with small pixels, the answer is easy - use subs that are 1/5 as long and have 5x as many of them.

DSLRs have well depth equivalent to CCDs, but they end up with white stars because users have wound up the ISO to get better looking subs. In the process, they reduce the effective well depth, throw away most of the dynamic range and end up with most stars saturated. High sky signal can also take over much of the remaining dynamic range, doing even more damage to the stars. This can be fixed by using low ISO, using short subs that keep the sky noise well to the left of the histogram, forgetting what the individual subs look like and concentrating on the final stack. However, new users fall into the trap of trying to get the subs to look nice - and they end with saturated stars.

Edit: It would be very helpful if you could post some images showing the difference between your two sensors.

Last edited by Shiraz; 03-06-2016 at 01:42 PM.
Reply With Quote
  #17  
Old 03-06-2016, 10:28 AM
Slawomir's Avatar
Slawomir (Suavi)
Registered User

Slawomir is offline
 
Join Date: Sep 2014
Location: North Queensland
Posts: 3,240
I think what might add to the impression that small pixels wreck stars more effectively than large pixels is the aperture of the telescope (given the same resolution).

Correct me if I am wrong, but from what have been said it seems that large aperture scope is more capable of producing tight stars than a telescope with small aperture. And since generally cameras with large pixels are put at the end of a telescope with a large aperture, it might seem that CCDs with large pixels have more effective anti-blooming and saturated pixels do not bleed charge as much to neighbouring pixels (not to mention that large telescope and large camera will be sitting on a well behaving mount, unlike most of small scopes with small CCDs). I believe this might be the main reason for nicer looking stars from large CCDs.
Reply With Quote
  #18  
Old 03-06-2016, 11:01 AM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by Slawomir View Post

large aperture scope is more capable of producing tight stars than a telescope with small aperture.

(not to mention that large telescope and large camera will be sitting on a well behaving mount, unlike most of small scopes with small CCDs). I believe this might be the main reason for nicer looking stars from large CCDs.
My understanding is that, once scopes get above about 200mm in Australian conditions, the atmosphere alone determines how tight the stars are, with the exception of the few very bright stars, where the diffraction/seeing skirt is tighter with larger scopes. However, in Australian seeing, the resolution of things such as galaxies etc will be pretty much the same for scopes of 200mm and up - a 200mm scope will do just as well as the AAT in resolving galaxy detail.

the above profiles take no account of mount performance, but this will have the effect of increasing the profile width for the best resolving setups, but not affecting lesser ones as much (if they already have broader profiles). So maybe mount noise will even the field the bit.

Last edited by Shiraz; 03-06-2016 at 01:43 PM.
Reply With Quote
  #19  
Old 03-06-2016, 12:08 PM
Slawomir's Avatar
Slawomir (Suavi)
Registered User

Slawomir is offline
 
Join Date: Sep 2014
Location: North Queensland
Posts: 3,240
Thanks Ray for confirming that. I believe that apart from a few notable exceptions, the majority of astroimagers who use cameras with tiny pixels have a scope with aperture less than 200mm
Reply With Quote
  #20  
Old 03-06-2016, 12:57 PM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
Another aspect that hasn't been mentioned yet is the impact of microlenses. KAF-8300 stars are pretty easy to pick once you're familiar with the "look." I have been unfairly accused of adding diffraction spikes to stars in refractor images taken with a KAF-8300! As well as an effect on star shape I believe they are responsible for an amount of bloat as well, presumably through scattering some of the incident light.

I don't know enough about sensor technology to be sure but it seems very likely that small pixels have a lower fill factor and place greater demands on physically smaller microlenses. I'll have to finish reading some of those big, fat books on CCD technology sitting on my bookshelf

Cheers,
Rick.
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 04:54 AM.

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