1.5.0 prerelease 2

[rthorvald]

A low value of Albedo can be used to eliminate the glow. I dunno if that'll do what you want, though. e.g.
Albedo 1e-32

Thanks for the tip; i??ll test...

- rthorvald

[chris]

With the latest CVS Celestia, here's Saturn and its ... moons or stars ?

I want to attempt to explain what's going on here . . .

First, the reason for adding glare halos is to give the appearance of brilliance to very bright objects. If you observe a light source in a dark room, you will observe that it appears to be surrounded by a glowing halo. Celestia exploits this effect in attempt to increase the apparent dynamic range of the computer monitor; otherwise, all objects above a particular intensity will all appear to be equally bright.

In older versions of Celestia, the halos were only applied to stars. This is unrealistic. If you're looking at the sky from Earth, Venus should appear brighter than any star; but, without the glare halo, Sirius and several other stars will seem to be the brighter objects.

The glare halos around planets and moons become a problem as the camera approaches these bodies and they transition from point sources to resolvable discs. For consistency, the glare halo should be applied to every point in the disc. In Cham's image, Saturn should be glowing brilliantly, and the interior of the disc should be washed out by the glare. But, nobody wants this (though I'll revisit the idea when I write more on HDR rendering.)

There needs to be a compromise someplace . . . Where should we sacrifice realism? I can attempt to tune the glare parameters somewhat, and get more acceptable images. The minimum visible magnitude is adjusted with the [ and ] keys. Perhaps we could have a separate control for the maximum magnitude representable? This 'saturation point' is when halos start to appear.

Here is a useful paper explaining the origin of glare effects in the human eye, and how they may be simulated in computer generated imagery:

http://www.graphics.cornell.edu/pubs/1995/SSZG95.pdf

--Chris

[selden]

The glare halos around planets and moons become a problem as the camera approaches these bodies and they transition from point sources to resolvable discs. For consistency, the glare halo should be applied to every point in the disc. In Cham's image, Saturn should be glowing brilliantly, and the interior of the disc should be washed out by the glare. But, nobody wants this [...]

I think this last statement may be due a misunderstanding.

I may be sounding like a broken record here, but I think that if glare is enabled (scaled disc mode), then it should be enabled for all bright objects, including washing out illuminated planetary and satellite surfaces. But I think that the user should be able to disable all glare (fuzzy star mode? = insert attenuation filter in viewing device or stop down aperture) in order to see surface features that otherwise would be invisible due to the blinding light.

I think that trying to satisfy both requirements (emulate illumination levels + be able to see dim features) at the same time is inappropriately unrealistic, both in appearance and in coding effort.

[Cham]

I say : remove completely this glare attempt. It's really a bad effect. It doesn't feel realistic at all. We never see something like this on pictures. And when the sky is very clear and black, I never see planets glowing like this on the real sky. To be honest, I'm about to trash Celestia if we come to something like that in Celestia.

[t00fri]

I want to attempt to explain what's going on here . . .

First, the reason for adding glare halos is to give the appearance of brilliance to very bright objects. If you observe a light source in a dark room, you will observe that it appears to be surrounded by a glowing halo.
...
--Chris

In my experience these kind of halos mainly occur in case that people's eyeglasses are badly cleaned...


In short: the above image does NOT at all create an impression of excessive brightness, but rather of 'fog in the Universe'...That's precisely how the stars look through my 8 inch Celestron telescope if the humidity exceeds 80%


Bye Fridger

[neo albireo]

I may be sounding like a broken record here, but I think that if glare is enabled (scaled disc mode), then it should be enabled for all bright objects, including washing out illuminated planetary and satellite surfaces. But I think that the user should be able to disable all glare (fuzzy star mode? = insert attenuation filter in viewing device or stop down aperture) in order to see surface features that otherwise would be invisible due to the blinding light.

I think that trying to satisfy both requirements (emulate illumination levels + be able to see dim features) at the same time is inappropriately unrealistic, both in appearance and in coding effort.

I totally agree.

[Buzz]

As I wrote before, the "glowing moons" don't work for me. This time I wanted to see some examples. An illustration of what Selden said can be seen on APOD: http://antwrp.gsfc.nasa.gov/apod/ap060622.html. Planets do have a halo here, but their interior is washed out with a uniform colour.

Buzz

PS: I would still very much like to see washed out stars!

[chris]

I may be sounding like a broken record here, but I think that if glare is enabled (scaled disc mode), then it should be enabled for all bright objects, including washing out illuminated planetary and satellite surfaces. But I think that the user should be able to disable all glare (fuzzy star mode? = insert attenuation filter in viewing device or stop down aperture) in order to see surface features that otherwise would be invisible due to the blinding light.

I think that trying to satisfy both requirements (emulate illumination levels + be able to see dim features) at the same time is inappropriately unrealistic, both in appearance and in coding effort.

But what about stars? Surely you want a glare effect to appear in some cases. If you're in the Solar System staring at the Sun, there needs to be some effect indicating its brilliance, regardless of what the star mode is. So one possibility is to only use the glare effect for extremely bright objects. This will make things look a little better, but it's only a partial solution.

--Chris

[chris]

I want to attempt to explain what's going on here . . .

First, the reason for adding glare halos is to give the appearance of brilliance to very bright objects. If you observe a light source in a dark room, you will observe that it appears to be surrounded by a glowing halo.
...
--Chris

In my experience these kind of halos mainly occur in case that people's eyeglasses are badly cleaned...

I don't wear eyeglasses, and when I stare directly at a bright light I can see a halo around it caused by scattering of the light within my eye.

In short: the above image does NOT at all create an impression of excessive brightness, but rather of 'fog in the Universe'...That's precisely how the stars look through my 8 inch Celestron telescope if the humidity exceeds 80%

I agree that the glare needs to be changed . . . Are you saying that it should be omitted completely? Or for just some objects?

--Chris

[t00fri]

...
I agree that the glare needs to be changed . . . Are you saying that it should be omitted completely? Or for just some objects?

--Chris

Let me get serious again.

It is correct that bright stars look precisely like in your above picture after my Schmidt corrector plate started fogging in at times of high humidity!

I agree that it is hard to render the huge, exponentially varying range of star brightness without "bad tricks".

The question seems to be what is the best one among the "bad tricks"?

It seems for now that most people (including myself) find the old stars or at best a "mild modification" of the old stars the better compromise. Since we did not discuss the PHYSICS of rendering stars yet, it's hard to make any predictions about what might be possible eventually. So far many people feel that your present attempts are not very convincing.

Bye Fridger

[t00fri]

Just to continue.

Chris,

you assumed particular behaviours of halo sizes, star core sizes, Gaussian width behaviour, clipping thresholds, .... when the star brightness increases.

None of these assumptions you did justify from a physics point of view. So what is the basis of your code? Without knowing this, it's really hard to form a concrete proposition for possible improvement.

Bye Fridger

[chris]

Just to continue.

Chris,

you assumed particular behaviours of halo sizes, star core sizes, Gaussian width behaviour, clipping thresholds, .... when the star brightness increases.

None of these assumptions you did justify from a physics point of view. So what is the basis of your code? Without knowing this, it's really hard to form a concrete proposition for possible improvement.

Bye Fridger

I don't have time to write in much detail right now . . .

Clipping:
A saturation brightness is chosen (somewhat arbitrarily) and intensities above this value are clipped. It's that simple. Whatever detector you choose will have some finite response range.

Gaussian FWHM:
It's constant, because the aperture is currently assumed to be constant. As far as I know, that's physically correct.

Halo size:
I picked a value that I thought looked good. This could be improved.

Star core size:
It's different in the various star rendering modes. In scaled disc mode, a linear detector (CCD-like) is assumed. The size of the star grows as the Gaussian PSF scales with the star's brightness.

Point and 'fuzzy disc' stars:
A perfect detector is assumed with point stars, though glare halos are still used. With fuzzy disc stars, the size is constant, but greater than one pixel. This roughly simulates the logarithmic response of the human eye, though really, the size of the star should increase slightly with brightness.

--Chris

[t00fri]

Just to continue.

Chris,

you assumed particular behaviours of halo sizes, star core sizes, Gaussian width behaviour, clipping thresholds, .... when the star brightness increases.

None of these assumptions you did justify from a physics point of view. So what is the basis of your code? Without knowing this, it's really hard to form a concrete proposition for possible improvement.

Bye Fridger

I don't have time to write in much detail right now . . .

Clipping:
A saturation brightness is chosen (somewhat arbitrarily) and intensities above this value are clipped. It's that simple. Whatever detector you choose will have some finite response range.

Gaussian FWHM:
It's constant, because the aperture is currently assumed to be constant. As far as I know, that's physically correct.

Halo size:
I picked a value that I thought looked good. This could be improved.

Star core size:
It's different in the various star rendering modes. In scaled disc mode, a linear detector (CCD-like) is assumed. The size of the star grows as the Gaussian PSF scales with the star's brightness.

Point and 'fuzzy disc' stars:
A perfect detector is assumed with point stars, though glare halos are still used. With fuzzy disc stars, the size is constant, but greater than one pixel. This roughly simulates the logarithmic response of the human eye, though really, the size of the star should increase slightly with brightness.

--Chris

Chris,

further up, I have tried to analyse various aspects of star rendering as they appear from a physicist point of view. Notably I have tried to separate various device dependent phenomena from basic and perfectly well-understood behaviours in optics. You did not comment how your various implicit assumptions fit into such a generally accepted framework.

So are we now leaving our hitherto agreed assumption that anything we see in Celestia is to correspond to naked eye vision? Once you subscribe to CCD response characteristics, also the spectral response needs to be considerably enlarged.

Or are we supposed to consider your present star modelling rather independent of serious physics constraints, mainly determined by subjective judgement?

Just let us know your general attitude.

Bye Fridger

[chris]

Chris,

further up, I have tried to analyse various aspects of star rendering as they appear from a physicist point of view. Notably I have tried to separate various device dependent phenomena from basic and perfectly well-understood behaviours in optics. You did not comment how your various implicit assumptions fit into such a generally accepted framework.

So are we now leaving our hitherto agreed assumption that anything we see in Celestia is to correspond to naked eye vision? Once you subscribe to CCD response characteristics, also the spectral response needs to be considerably enlarged.

As I stated, only the scaled discs mode attempts to model anything like CCD response characteristics. However, switching to scaled discs mode does not change the rendering of planets or other objects. Nor is the spectral response altered at all; that's simply not a feature that I've sought to emulate.

Or are we supposed to consider your present star modeling rather independent of serious physics constraints, mainly determined by subjective judgement?

It's partly determined by physics, but there's also a subjective element to it, because adhering strictly to a physically based approach produces nonsensical looking results (at least without an HDR frame buffer.)

--Chris

[chris]

I say : remove completely this glare attempt. It's really a bad effect. It doesn't feel realistic at all. We never see something like this on pictures.

Don't we?



To be clear, I'm not arguing against changing or eliminating the halos on planets; I'm just challenging you to reconsider your views on what is and is not realistic.

--Chris

[t00fri]

Or are we supposed to consider your present star modeling rather independent of serious physics constraints, mainly determined by subjective judgement?

It's partly determined by physics, but there's also a subjective element to it, because adhering strictly to a physically based approach produces nonsensical looking results (at least without an HDR frame buffer.)

--Chris

I am not at all convinced that this is a correct statement. I agree that with HDR frame buffers there is more opportunities for getting a really nice result, but if the underlying physics is incorrect, HDR doesn't help either.

Without HDR, one crucial issue would be how the ratio of core size to halo size is supposed to behave with increasing nominal star brightness.

You simply assumed something which you did not justify with physical arguments.

I could argue if this were a physics discussion and not a subjective one, that the halo should grow less strong while the core size should grow stronger than in your modelling. And so on...

My statement is that we can predict these behaviours from theory unless the empirical detector device characteristics are such as to wash out almost everything we know from basic theory..

Bye Fridger

[chris]

Or are we supposed to consider your present star modeling rather independent of serious physics constraints, mainly determined by subjective judgement?

It's partly determined by physics, but there's also a subjective element to it, because adhering strictly to a physically based approach produces nonsensical looking results (at least without an HDR frame buffer.)

--Chris

I am not at all convinced that this is a correct statement. I agree that with HDR frame buffers there is more opportunities for getting a really nice result, but if the underlying physics is incorrect, HDR doesn't help either.

Well, I agree with the last part completely, and I think that we should strive to be more rigorous when we move to HDR. And then the results from that work can be used to make more informed choices on rendering stars (and other objects) in the non-HDR paths.

Without HDR, one crucial issue would be how the ratio of core size to halo size is supposed to behave with increasing nominal star brightness.

You simply assumed something which you did not justify with physical arguments.

Well, I'm still doing research on glare. The paper I linked to earlier provides a lot of information on the causes and appearance of glare in the human eye:

http://www.graphics.cornell.edu/pubs/1995/SSZG95.pdf

I could argue if this were a physics discussion and not a subjective one, that the halo should grow less strong while the core size should grow stronger than in your modelling. And so on...

I'd like to hear your argument, though I think the paper by Spencer et al already has me convinced--for the human eye, that is; I'm less informed about glare effects due to scattering in telescope and camera optics.

--Chris

[t00fri]

Chris,

that's certainly an interesting & relevant paper. I'll get back to you once I found time to read it carefully.

Bye Fridger

[PlutonianEmpire]

For the record, what's HDR? High Definition Rendering?

[chris]

For the record, what's HDR? High Definition Rendering?

HDR = High Dynamic Range. Essentially, it's the ability to represent a large range of brightnesses in an image.

--Chris