About visibility of stars

[dirkpitt]

And actually, if you looked at a star from that close, you'd probably be permanently blinded.

You'd also be vaporized I'd imagine...

[Malenfant]

"Better" is a matter of opinion, I guess. I'm more interested in Celestia as a visualization tool than as a visual simulation.

I'd like Celestia to answer a simple question for me - what do other stars look like? We see a glaringly bright yellow ball in our sky, but what would you see if we were close enough to an M V star for it to be the same size as the sun in our sky? Would it still be blindingly bright? Would it look visibly red? And what about a red supergiant? Or a star like Sirius?

Certainly right now the way Celestia renders stars is not accurate. Making it more accurate, so that it could answer questions like the one I asked here, would make it a very useful educational tool, simply because people generally don't stop and think about how different (or similar?) things may look compared to our own terrestrial environment.

[ElChristou]

Guys, apparently Chris is working on the stars rendering, so better wait for news; now the question of visibility of stars when a major body is present on screen present also a problem: if the observer pass in the dark side, the stars will have to gain again in brightness, right? so finally this kind of fluctuation will not be so easy to calculate, I imagine...

[dirkpitt]

Here is a paper on perceptual tone mapping. There're many ways to show high dynamic range (superbright/overdark) imagery on a
monitor and this paper discusses ways to map out-of-range values to displayable ones, according to a human eye perceptual model.
Glare is one of the possible effects, but there are others. Being able to switch on/off HDR rendering in Celestia would be important.

Another easier article by Anandtech is here. There's a nice picture showing the sun gradually saturating the entire frame as it
comes out from behind the clouds.

[ElChristou]

HDR would be fantastic, but then Celestia won't become and elitist soft? (working fine only on hight level cards?)

[selden]

Chris has frequently added eye-candy that doesn't work on older cards, but they've been additions. Older cards show no less than they always did. The shadows of rings, for example: the shadows cast by rings onto the planet can't be seen with MX cards in the OpenGL Vertex Program render path. (I don't know if they're visible in the OpenGL 2.0 render path on MX cards.)

[buggs_moran]

I'd like Celestia to answer a simple question for me - what do other stars look like? We see a glaringly bright yellow ball in our sky, but what would you see if we were close enough to an M V star for it to be the same size as the sun in our sky? Would it still be blindingly bright? Would it look visibly red? And what about a red supergiant? Or a star like Sirius?

It's my understanding that they would still look white to our eyes when unfiltered/unscattered by an atmosphere or other matter. This is the same way that the sun looks white to the astronauts. It is due to the fact that even though stars may be shifted one way or the other in the spectrum, the light is spread across the entire spectrum so our eye receptors see all the colors. The true problem is the brightness of the stars. From a certain distance, your eye receptors get overloaded by all the colors, so you see white. Move away from the star and let some of the photons fall off (scattering due to distance and filtering due to matter) so not so many are entering your eye and you get the reds of Arcturus, Antares, Betelgeuse and the blues of Rigel and Sirius et cetera. On extrasolar planets, strange colors would be more possible with thicker atmospheres. Ours scatters blue (and thankfully many UV, x-rays and gamma rays), and allows us to see a (red/green) yellow sun. I don't know much past 1st year college Chemistry (and that was a long time ago), so light scattering due to atmosphereic makeup is not something I can explain.

[Christophe]

Just for info, Chris is working on HDR rendering for Celestia and achieving a more realistic human vision simulation is one of his goals.

[dirkpitt]

HDR would be fantastic, but then Celestia won't become and elitist soft? (working fine only on high level cards?)

Even my 2-year old 9700 just barely runs any "realtime" HDR demos I throw at it. Hopefully Celestia will be optimized to continue to run at a reasonable frame rate even for the "proletariat".

[Malenfant]

What is this HDR that keeps being mentioned here? And what would it actually mean for star rendering?

[hank]

What is this HDR that keeps being mentioned here?

Take a look at the articles dirkpitt referenced above.

- Hank

[ElChristou]

Guys, I been thinking in this HDR topic... and I want to give here my little POV; it's a really simple analyse, based on what I can see on screen, I'm not a specialist of 3D real time rendering, so if I'm saying a big stupidity please let me know quickly...

HDR mean a work on contrast to simulate the reaction of our eyes when we pass form a dark ambient to a bright one or the contrary. In the case of Celestia, apart the sudden apparition of a body in the screen (in this case the stars will fade to invisible), this effect will not be very useful. HDR mean also simulating glare effects; this is fine in a complex environment, but in our case, we are dealing with an almost pure back environment... so the question is: how to contrast more than black vs white? this is not possible of course...

So if i'm not wrong, we have 2 principal problems: visibility of stars versus major bodies and glare effect on stars.

For the glare effect, IMO this can be achieved right now without rewriting the Celestia engine.
For the visibility of stars, I want to ask the dev team (and particularly to Chris of course) if there is not some simplest solution more accessible to all render paths to simulate this?
HRD mean rewriting the engine, mean cost in calculation, mean fall of the fps... mean better cards; as Celestia is a knowledge tool (used in education), using this kind of ultimate technology for realtime 3D rendering is really necessary? (not every school and college have the last 3D cards in their equipment...)

Of course I understand perfectly the interest in using such a cool tool at the dev level, but a balance should be done; I almost sure some more simple tricks can be found, not so heavy to implement and perhaps more accessible to the basic user...

[Malenfant]

HDR mean a work on contrast to simulate the reaction of our eyes when we pass form a dark ambient to a bright one or the contrary. In the case of Celestia, apart the sudden apparition of a body in the screen (in this case the stars will fade to invisible), this effect will not be very useful. HDR mean also simulating glare effects; this is fine in a complex environment, but in our case, we are dealing with an almost pure back environment... so the question is: how to contrast more than black vs white? this is not possible of course...

So if i'm not wrong, we have 2 principal problems: visibility of stars versus major bodies and glare effect on stars.

I'm not after "useful", I'm after "realistic". And both these effects would make Celestia much more realistic as a means to represent what we'll actually SEE in space.

The glare effect is important because if we see the sun and the earth in the same view, then darn straight we won't be able to see anything on the side of the earth facing us because the sun is so bright. Only when it disappears behind the planet will we start to see something (and then only if the moon happens to be lighting up the surface, or if there are nightlights). But there should be a few seconds when our "eyes" adapt to the new low light level first.

As for losing the background stars when bright objects are in the view, well, that's what happens in such situations. If Celestia is to be realistic in other ways, why not that too?

Ideally, an option to just turn on or off these glare/HDR functions would be nice. That way both sides (more realistic vs more simulationist) would be pleased.

And I can tell you from experience with EVE that it is MUCH more aesthetically pleasing to see the glare and so on than it is to just have what Celestia currently shows. It just looks more real, and more breathtaking and impressive.


One other point is that we'll never actually 'escape' Lens Flare - if you're in a spaceship then you'd be looking out of a window, which would refract/diffract the light. If you were in a space suit, your visor would do the same. And even if you were able to survive in space without a spacesuit, the fluid in your eyes would do the same thing. So it's actually more realistic to include SOME degree of lens flare than none at all (doesn't have to be excessive, but it should be present at least as an option).

[ElChristou]

Malenfant, we are both agree to say that Celestia must simulate some more realistic effects; the dev team is conscious of that, Chris is working on some solutions, perhaps with what propose the HDR, and to tell the truth it's almost perfect for what we want.
I'm just wondering if it's not using a too complex and heavy tool (normally dedicated to modify complex scenery) for what we have to simulate in Celestia: fading a background and adding some glare on stars...

[Malenfant]

Malenfant, we are both agree to say that Celestia must simulate some more realistic effects; the dev team is conscious of that, Chris is working on some solutions, perhaps with what propose the HDR, and to tell the truth it's almost perfect for what we want.
I'm just wondering if it's not using a too complex and heavy tool (normally dedicated to modify complex scenery) for what we have to simulate in Celestia: fading a background and adding some glare on stars...

Well it may not be as complex as you think. Most renderers have to cope with issues like radiosity and ambient light and so on, bouncing off many other objects in a scene.

In Celestia, all we have to worry about is the star and any nearby reflective or emissive light sources (ie planets, asteroids etc). The only radiosity we need to worry about is light reflected off a moon onto a planet and vice versa (or any other objects close enough to be lit by light reflected off a planetary body).

In other words, in all likelihood we could use a much simplified but still effective version of HDR for Celestia.

[ElChristou]

...In Celestia, all we have to worry about is the star and any nearby reflective or emissive light sources (ie planets, asteroids etc). The only radiosity we need to worry about is light reflected off a moon onto a planet and vice versa (or any other objects close enough to be lit by light reflected off a planetary body).

HDR as nothing to do with radiosity, it's a modification of the contrast of an already calculate frame to create the effect of "even much brighter" or "even much dark"... Dikpitt docs explain much better than I the concept...

In other words, in all likelihood we could use a much simplified but still effective version of HDR for Celestia.

It's my question...

[Telepath]

The only radiosity we need to worry about is light reflected off a moon onto a planet and vice versa (or any other objects close enough to be lit by light reflected off a planetary body).

What about light reflected off models? Take fsgregs Space Station V as an example. In reality, every spoke and surface would reflect light (and shadows) onto the other parts of the model, or a nearby model, or a nearby asteroid,planet, or whatever.
You can't only consider Planets, Moons, etc. if the aim is maximum realism. In this respect, Celestia is no different than any other 3D simulation, the challenges and complexities are the same, and I believe any 'simplification' or compromize in the rendering department will involve a loss of realism.
At the moment Celestia doesn't even do shadow casting by models, I'd rather see priorities on that, ahead of radiosity.

Problems with transparent models is also quite a serious problem in some situations. Much as I agree with the sentiment of maximum reality for future development, I'd rather wait for things like radiosity, and see the developers focus on serious problems like this transparency issue first. There's not much point having a beautifully rendered and reflecting model, if you can see straight through parts of it that you shouldn't!

Of course, everyone's going to have a different set of priorities and uses for Celestia. I'm sure the planetary scientists out there are not too worried about problems with rendering of models, yet the educationalists are not in a position to keep up with high-spec graphics requirements, and I imagine (and hope) that Chris and the other developers are constantly trying to balance these needs.

Personally, I think the ability to extend the Celestia universe with 3rd party models is a great capability, which shouldn't nescessary take a back seat to planets, etc. when considering future directions and priorities. (Just look at some of the excellent addons on the motherlode)

Just my 2 cents.

[ElChristou]

Telepath, what you are proposing is full raytracing or even radiosity rendering, which is not possible in 3D real time because of the calculation power needed to do such rendering... let's wait 20 years to see that

[ElChristou]

WARNING, artistic views only...

Guys, for fun, some others simulation of the topic:

top: what we have now.
middle: the sun is so bright that our eyes cannot perceive dimmest sources of lights (stars, night lights)
bottom: the same + a PS generated Lens Flare (105 mm)



just an illustration of what HDR is suppose to do:
Let's say you are doing a rotation, suddenly the earth enter the screen:

top: during 1 or 2 seconds the earth will be over contrasted; then will fade in 1or 2 seconds more to
middle: the earth is normally contrasted (normal rendering), but stars in the background won't be visible or perhaps only the brightest once...
bottom: what we have now.



What we must keep in mind is the movement; in the first example, once the earth occult the sun, stars and night lights will appear...

[Telepath]

Telepath, what you are proposing is full raytracing or even radiosity rendering, which is not possible in 3D real time because of the calculation power needed to do such rendering... let's wait 20 years to see that

I agree.
Actually, I was proposing they **not** waste time on it

I'd rather ..... see the developers focus on serious problems like this transparency issue first