Archived

This topic is now archived and is closed to further replies.

Cipher3D

is it really possible to achieve total photorealism?

Recommended Posts

Cipher3D    340
for the last 10 years people have claimed one game engine or another were "frighteningly real" (DOOM, Quake, Quake II, Quake III). I believe this will continue on into far of the forseeable future. Is it trully possible to achieve photorealism? Even if we had a gigantic computer that rivalled even the one that ran The Matrix, we still have Quantum Mechanics to deal with - we can''t really determine the position of each Photon, and until we figure out how the world REALLY interacts (in the RW, there is no seperate graphics module, no seperate phsyics module - everything is tied together), we will be stuck approximating it. Any ideas?

Share this post


Link to post
Share on other sites
Leffe    301
Temporary photorealism should be quite possible with the right equipment.

Permanent, a.k.a. Teh Matrix, I don''t think so.

By temporary I mean just taking a quick look, not staring at the render for hours trying to see something that looks unreal.

Of course, this is just for a still image, not a realtime render. Motion is where you can really see a difference. And you''ll quickly notice that it''s not real when there''s no movement. Quite a problem.

And no other stuff, just vision, ...

Share this post


Link to post
Share on other sites
Rocket05    152
there are ''levels'' of photorealism. To someone with bad eyesight, the original quake was probably quite realistic. As realism and sharpness in graphics increase, it fools a sharper and sharper eye. Although I dont know if we could ever really perfectly model and render the optical world, we can produce images of such sharpness and quality that we can fool even the sharpest eyes. Pre-rendered stills have already gotten to this point in some instances, there are a few images on the internet ray tracing competition website (www.irtc.com) that unless you didn''t know better, you''d probably believe that they were photographs.

http://www.irtc.org/ftp/pub/stills/2003-04-30/incubus_.jpg -- This one in particular stood out as one of the most realistic renderings i''ve ever seen. I''ve linked a few friends to that rendering and told them it was a picture i took of an apartment, and they were easily fooled, they had no idea its not a real picture.

I believe it only a matter of time before we see such realism in real-time.

Share this post


Link to post
Share on other sites
Raloth    379
quote:
Original post by Cipher3D
for the last 10 years people have claimed one game engine or another were "frighteningly real" (DOOM, Quake, Quake II, Quake III).

I believe this will continue on into far of the forseeable future.

Is it trully possible to achieve photorealism? Even if we had a gigantic computer that rivalled even the one that ran The Matrix, we still have Quantum Mechanics to deal with - we can''t really determine the position of each Photon, and until we figure out how the world REALLY interacts (in the RW, there is no seperate graphics module, no seperate phsyics module - everything is tied together), we will be stuck approximating it.

Any ideas?
Sure we can... Just model string theory .

Share this post


Link to post
Share on other sites
Dreamforger    122
Of course we will. At first only with very constrained environments, later with more complex ones until we can no longer discern from reality. We always were very good at displaying a unlit, windowless room (black) - don''t laugh it''s just a very powerfull contraint. At the moment I''d say water is a problem, volumetric structures like fog, gas are being conquered as we speak.

But there are some psychological aspects you have to consider. As long as we are bound to that pathetic flat surface called monitor (which as the name implies only lets you monitor, which is far weaker than view), our mind has a very easy job finding reasons why this is not real.

So photorealism is not only a question of the available processing power, data and alogrithms. Our monitors can not display every color, currently used color-models cannot describe all natural colors.

Startrek''s holodeck is still a long way off, but we will arrive there one day. But not in my days and I''m not that old yet.
at least I''m positive that some kind of volumetric viewer will be available within the next 10 years

---------------------------
I may be getting older, but I refuse to grow up

Share this post


Link to post
Share on other sites
rohde    432
At some point in time I do believe it would be possible.

Hardware that raytraces for instance would go along way, but there is still a long way, especially if you want to see "photorealistic" humans (but thanks to pioneer works by LucasArts refined by the people behind LOTR CGI lighting scheme we are on the way).



Human beings, by changing the inner attitudes of their minds, can change the outer aspects of their lives.

William James (1842 - 1910)

Share this post


Link to post
Share on other sites
Dreamforger    122
quote:
Original post by Rocket05
http://www.irtc.org/ftp/pub/stills/2003-04-30/incubus_.jpg --

Thank you for making my point. If you hadn''t wold me that this image had been rendered I might have been fooled. But since you told us, I can''t get my mind to accept the image as a photograph anymore. Some tiny aspects just feel wrong, and I can''t put my finger on most of them.

Share this post


Link to post
Share on other sites
rohde    432
quote:
Original post by Dreamforger
Startrek''s holodeck...

Damn, that would be sweet to live to experience that.





Human beings, by changing the inner attitudes of their minds, can change the outer aspects of their lives.

William James (1842 - 1910)

Share this post


Link to post
Share on other sites
superpig    1825
I agree that it''s practically impossible to model every photon in a system, but is that really necessary? The human eye can''t distinguish single photons; they don''t tend to happen on their own anyway.

Our eyes do have maximum resolutions. Try this little experiment: Draw two dots of equal size on a piece of paper, about a centimeter apart. Stick it up on the wall, and walk backwards away from it. At some point, you won''t be able to tell the two dots apart. That''s your resolution; the light from each dot is falling on the same cells in the retina.

Taking Rocket05''s picture as an example, standing away from it makes it more believable, I feel. The suggestion has been made the the brain just ''guesses'' at what it can''t actually make out, and when the rest of the picture is so close to photorealistic, the brain gives it the benefit of the doubt and provides the rest.

So the first requisite for photorealism will be monitors and video cards which can operate at resolutions beyond that of the human eye (at the average computer user''s distance from the screen), IMO.

Richard "Superpig" Fine
- saving pigs from untimely fates, and when he''s not doing that, runs The Binary Refinery.
Enginuity1 | Enginuity2 | Enginuity3 | Enginuity4 | Enginuity5
ry. .ibu cy. .y''ybu. .abu ry. dy. "sy. .ubu py. .ebu ry. py. .ibu gy." fy. .ibu ny. .ebu
"Don''t document your code; code your documentation." -me
"I''m not a complete idiot; parts of me are missing." -}}i{{

Share this post


Link to post
Share on other sites
Cipher3D    340
I am nothing short but blown away from that incubus picture...hell...how do they DO THAT? it must take them a friggin long time just to model everything....

Share this post


Link to post
Share on other sites
cowsarenotevil    3006
Assuming we''re aiming for digital photo quality, it can already be done. The only thing really lacking so far is lighting, because it''s very difficult to model surfaces well, and sunlight is very complex. But with pixel shaders and stuff, I don''t think photorealism is that far off.

Share this post


Link to post
Share on other sites
lc_overlord    436
The lighting in that image is good but not that good.
First there is the lack of caustic''s.
Second, sunlight is all wrong, normaly it''s much more powerfull.
Third, everything reflect''s way to mutch light, it''s like they used linear falloff.

IT''s good, but not that good.

But if you only look at the realism factor in id''s doom and quake series games you can extrapolate the realism factor in future games.
So "photorealism" is not that far away.

---------------------------------
For an overdose of l33tness, flashbang.nu

Share this post


Link to post
Share on other sites
dmounty    122
I think one problem with that photo is the texture of the surfaces in general... teh look a little too clean. If they all had a little more texture to them.. it would be considerably more convincing. Also the plant models look a little strange. On balance though.. I think we can be almost entirely certain that in time entirely photo realistic rendering in realtime will be available. If you look at a correctly globally illuminated model of a simple object on a flat plane (eg a table) it can look entirely convincing. Just look at a sphere on a plane rendered with a global illumination scheme (ie photon mapping + montecarlo raytracing). A rendering scheme of that sort is capable of photorealistic rendering.. but it requires a more detailed description of all the surfaces involved. To get the best possible lighting.. an extended environment must be provided as well... to make the lighting appear more natural. I suspect that with another 5 years of research into such techniques... individual frames that appear photographic may be rendered. Another 5 years, and it'll be availabe realtime on super computers.. another 10 years.. and it'll be on desktops.. I suspect that realtime gaming systems will have photorealistic graphics in 20 years time. I think this is more that feasible, and will be fun .

EDIT: with regards to that horse... it looks real to me, because of the level of detail, and apparently random(ish) formation of the hairs on it's mane.. as well as the texture of it's "fur". Also, the transitions in colour around the nose look very convincing. The depth of field is also a strong visual clue (ie DoF-less images tend to appear far less convincing). Also.. once we get photo realism, what next... well... we go beyond reality. We provide access to scales, locations and environments that are entirely alien and unfamiliar to us.

[edited by - dmounty on January 18, 2004 3:48:07 PM]

Share this post


Link to post
Share on other sites
Guest Anonymous Poster   
Guest Anonymous Poster
"Original post by Rocket05
http://www.irtc.org/ftp/pub/stills/2003-04-30/incubus_.jpg"

I immediately see this is not a real picture. I can''t believe anyone would see this as real photograph. Yes, maybe when you see it from very far away like someone mentioned. But that''s not fair, because you don''t see any details.
There are however images where I was also fooled. I think there is a website which is something like "fake or not?" and you have to say from 10 images if they are CG or real. And then you most likely find out you had some wrong

Share this post


Link to post
Share on other sites
bishop_pass    109
quote:
Original post by dmounty
EDIT: with regards to that horse... it looks real to me, because of the level of detail, and apparently random(ish) formation of the hairs on it''s mane.. as well as the texture of it''s "fur". Also, the transitions in colour around the nose look very convincing. The depth of field is also a strong visual clue (ie DoF-less images tend to appear far less convincing).
And all of those things are things to strive for in computer generated imagery. You need to study the subtle little details and attempt to reproduce them. Nobody really ever goes that far.

In the living room image, where''s a fabric discoloration on the couch. No, I don''t mean one that is necessarily obviously visible. Where are the dust motes? Where is the dust in the corners? Where is a boot scuff on the bottom edge of the wall? Why isn''t there the slightest hint of a paint dribble on the window frames?

Take a look through a home or architectural magazine and look at the photos. Very beautiful homes have tiny imperfections.

Share this post


Link to post
Share on other sites
Guest Anonymous Poster   
Guest Anonymous Poster
"Bishop: So is this a rendered image or not? And how can you tell? "

Real. I think the hair looks too good to be CG

Share this post


Link to post
Share on other sites
quasar3d    814
quote:
Original post by Anonymous Poster
There are however images where I was also fooled. I think there is a website which is something like "fake or not?" and you have to say from 10 images if they are CG or real. And then you most likely find out you had some wrong


I believe there was a test on alias wavefront''s image, but it were all very strange pictures anyway. Like a very close close up of a fork, or a two grass blades etc. I''ve looked at alias wavefront''s site, but I couldn''t find it any more.


My Site

Share this post


Link to post
Share on other sites
uutee    142
Well, I'll discuss my view from the perspective of a graphics *PROGRAMMER*...

Because even though good graphics Algorithms could indeed produce total photorealism (in theory), the weakest link can often be the modeling: it's hard to get all the parameters right and add enough realistic details.

>>we still have Quantum Mechanics to deal with - we can't
>>really determine the position of each Photon, and until
>>we figure out how the world REALLY interacts

As superpig pointed out, this isn't really even necessary, because surfaces and medias usually give *enough* realism (or should I say, leave very little error) to make images indistinguishable from photographs.

>>The only thing really lacking so far is lighting, because
>>it's very difficult to model surfaces well, and
>>sunlight is very complex.

Yes - from the algorithmic side lighting (this includes shadows) is the slowest thing and is a totally offline procedure if made ultimately realistic without hacks.

Just imagine the numerous ways in which light reflects in a simple cornell box with a diffuse and a specular objects filled with smoke - the lighting complexity simply is huge.

For lighting *interaction* in the *scene* we need to know something about the *scene* before we can model the interaction. For this I see render-to-texture a promising area for future techniques, as it's a fast and intuitive method (although the results can be a little erroreus.) Unfortunately even current graphics cards can't (to my knowledge) make too many render-to-textures per frame while retaining real-time speeds :/

>>I suspect that with another 5 years of research into
>>such techniques... individual frames that appear
>>photographic may be rendered. Another 5 years, and
>>it'll be availabe realtime on super computers..
>>another 10 years.. and it'll be on desktops..
>>I suspect that realtime gaming systems will
>>have photorealistic graphics in 20 years time.

Now *that's* what I call a realistic prediction of future.

- Mikko Kauppila

[edited by - uutee on January 18, 2004 5:23:26 PM]

Share this post


Link to post
Share on other sites
Mayrel    348
quote:
Original post by Cipher3D
for the last 10 years people have claimed one game engine or another were "frighteningly real" (DOOM, Quake, Quake II, Quake III).

I believe this will continue on into far of the forseeable future.

Is it trully possible to achieve photorealism? Even if we had a gigantic computer that rivalled even the one that ran The Matrix, we still have Quantum Mechanics to deal with - we can''t really determine the position of each Photon, and until we figure out how the world REALLY interacts (in the RW, there is no seperate graphics module, no seperate phsyics module - everything is tied together), we will be stuck approximating it.

Any ideas?


Our brain doesn''t distinguish between individual photons, so I see no reason why a renderer should need to do so. We only need to make something is that good enough to trick the brain into thinking that what it sees is real.

As it happens, visual realism is not that much of an issue for motion pictures. The things that are difficult to render, such as hair, are difficult not because of how they look , but because of how they behave .

CoV

Share this post


Link to post
Share on other sites
sBibi    241
this picture:

http://www.amazonsoul.com/dick04.jpg

_could_ look like a photograph, but there are some small details, I can't precisely tell witch ones, that make it look like a CG image.
(and btw, sometimes, real photographs do look like CG, or at least.. "unrealistic")

edit: perhaps it's the skin tint :|

[edited by - sBibi on January 18, 2004 7:47:04 PM]

Share this post


Link to post
Share on other sites