Digital Foundry Article Technical Discussion Archive [2010]

[Silent_Buddha]

Whole new way to think.

No, it's not. I've been playing around with stereoscopic games, movies and images since the mid 90's, although I haven't been as active in it the past few years. The basic concept and most of the technology hasn't changed all that much. Only the scale of displays, and more advanced shutter glasses (along with synching of frames) has. It's not that difficult to get your mind around.

Regards,
SB

 

[ihamoitc2005]

I still think you're exaggerating the differences. 3D is all about the fact that each eye gets the same information, but with just a slight enough difference so that it can perceive 3D. Had the information been completely different, it wouldn't have been able to perceive depth.

Of course there is a lot more going on in actual 3D spatial awareness, but just seeing distance accurately as simulated by the 3DTV technology that we're discussing now isn't that complicated, and you're turning it into something more than it actually is.

No it is different information. If it is same information it would be 2D.

Left Eye Frame (low res, text is pixel)

KLJHKHNAASREKSLAHGAR
OIUSDGLKQLKNASOIUIOOG
OQWQRLKSUHIGOIAOIERW
QWRQLKASIGOAIYUEQKKL

Right Eye Frame (low res, text is pixel)

XFASLKHGLKASRLAKGLAR
QWROIHSAALKNETOIHASO
CAJQROIAJHOASFOIAISHO
RWQOIJOASUFUAGJKBREI

See? Different information.

Small object in front-left of total view (brain receives 4 pixels)

Left Eye/Right Eye

O/C
Q/R

If same object is in 2D screen, the brain receives the same number of pixels (4) except both eyes gets identical pixels so there is no parallax so brain does not receive depth information. The texture is a little sharper in 2D (because all pixels are perfectly adjacent) but perception of geometry, lighting, depth, etc is much less.


@SB. Haha, my friend, I bet you have not had this. (Viewmaster)

http://www.amazon.com/Classic-Model-ViewMaster-Viewer-Two-Tone/dp/B001C1PCYU

 

[Silent_Buddha]

@SB. Haha, my friend, I bet you have not had this.

http://www.amazon.com/Classic-Model-ViewMaster-Viewer-Two-Tone/dp/B001C1PCYU

Hehe, had one of those as a little kid. I have owned various systems using shutter glasses as well as various glasses and head mounted units with LCD displays for each eye however. And have used virtually everything else except for some prototypes of systems that project the image directly onto your retina.

I did have an opportunity once at the University of Washington to try out one of the systems that project onto the retina, but never had the time to make it to the lab.

Regards,
SB

 

[TrungGap]

No it is different information. If it is same information it would be 2D.

Left Eye Frame (low res, text is pixel)

KLJHKHNAASREKSLAHGAR
OIUSDGLKQLKNASOIUIOOG
OQWQRLKSUHIGOIAOIERW
QWRQLKASIGOAIYUEQKKL

Right Eye Frame (low res, text is pixel)

XFASLKHGLKASRLAKGLAR
QWROIHSAALKNETOIHASO
CAJQROIAJHOASFOIAISHO
RWQOIJOASUFUAGJKBREI

See? Different information.

Small object in front-left of total view (brain receives 4 pixels)

Left Eye/Right Eye

O/C
Q/R

If same object is in 2D screen, the brain receives the same number of pixels (4) except both eyes gets identical pixels so there is no parallax so brain does not receive depth information. The texture is a little sharper in 2D (because all pixels are perfectly adjacent) but perception of geometry, lighting, depth, etc is much less.

No one is arguing that you're not seeing different information. It's just that those information are translated in your brain as depth not as resolution. Just as a game that runs at 30fps with resolution of 200x100 pixel vs 60fps with 100x100 pixel. Clearly the 60fps is delivering the same number of pixel per second, but it doesn't mean it perceived as resolution, but better fluidity.

 

[ihamoitc2005]

No one is arguing that you're not seeing different information. It's just that those information are translated in your brain as depth not as resolution. Just as a game that runs at 30fps with resolution of 200x100 pixel vs 60fps with 100x100 pixel. Clearly the 60fps is delivering the same number of pixel per second, but it doesn't mean it perceived as resolution, but better fluidity.

It is not pixels per second that counts but pixels per moment.

At anyone moment, 100x100x60fps is delivering half the pixels of 200x100 pixels. That is why it is perceived as lower resolution. Maybe if you "wobble" the picture at very high rate so you are rendering parallel details you can simulate single higher resolution image.

For 100x100x2 3D, only texture resolution appears less than 200x100x1 2D. This is because in 2D, all pixels are perfectly adjacent but in 3D, sometimes there is a large gap (front of scene) and in back of scene there is some overlap.

Information about lighting, object geometry, and scene depth is infinitely more in 3D.

So traditional idea of "resolution" cannot apply to this kind of perception.

Maybe you can think like this for easy 2D to 2D comparison.

One screenshot has double resolution textures for all objects and details like stones on the road are just drawn texture without any bump map technique. Other screenshot has less resolution textures for most objects but has parallax maps, etc... for nice round stones and other details. Which one gives you more detail?

 

[Silent_Buddha]

Once you actually use it you'll realise that say 200x100 pixels for each eye will equal 200x100 pixels in motion.

Everything you're saying may sound reasonable to yourself and possibly to others if you just say it enough in enough different ways, but once you see it, it's just 200x100 pixels.

It doesn't automagically appear to be 2x (or as your argument goes 4x) the resolution. Nothing is doubled.

It doesn't suddenly appear to be 400x100 or 200x200 or 400x200. It will still look exactly like 200x100.

Everything you say flies completely in the face of everyones experience with these systems.

Regards,
SB

 

[TrungGap]

So traditional idea of "resolution" cannot apply to this kind of perception.

So are you not arguing that resolution will be doubled? But just that there will be more information? I don't think anyone here disagree with that. They're just saying the resolution will not be accumulative for each eyes, but what you're getting is the depth part of the picture.

Here's a good test to try out. If you have a screen door (or mini blinds) taken in high resolution (2D) vs a lower resolution (3D). If you can't resolve the details of the screen door to the level of the 2D image, then you have not achieve the same resolution as the 2D with the 3D.

 

[ihamoitc2005]

Not what I am saying

So are you not arguing that resolution will be doubled? But just that there will be more information? I don't think anyone here disagree with that. They're just saying the resolution will not be accumulative for each eyes, but what you're getting is the depth part of the picture.

Sorry, my friend, this is not what I am saying.

Graphics resolution is not just about textures. Much more to 3D perception than "depth."

 

[Arwin]

Graphics resolution is not just about textures. Much more to 3D perception than "depth."

You have a very clear opinion on this, but do you perhaps have any cites to actual research that proves this?

 

[ihamoitc2005]

Binocular vision

You have a very clear opinion on this, but do you perhaps have any cites to actual research that proves this?

Good morning. Benefits of binocular vision (apes/monkeys, most predators) vs panoramic (most prey creatures, cows, deer. and most fish see) is well documented and researched.

I have to qualify to say that even most fish have tiny area of FOV that is binocular. We have most of our FOV as binocular. So we can perceive speed, distance, shapes, light/shadow/particle volume, materials (due to parallax view of reflecting light) etc. much better.

 

[dumbo11]

You have a very clear opinion on this, but do you perhaps have any cites to actual research that proves this?

I think the underlying effect is something like the term 'eye dominance'?

We get vision from 2 eyes, use them for depth perception - but most people (normally) only see the image through their dominant eye. So a 200x100 3d image would probably make us 'see' 200x100.

I think that's the best answer "mostly".

--

However, it seems likely that we do use binocular 'hints' to help us perceive the world, beyond depth perception.

Here's a paper on glossiness, which suggests that for 'very glossy' materials we use binocular vision, I believe there are others on other aspects of material identification:
http://www.journalofvision.org/content/4/9/4.full

There are some truly outlandish/crazy options:
- if you know that the right eye is dominant, could you cheat with the left eye's image a bit without the viewer noticing?
- could any part of the rendering be omitted/reduced, and rely on the viewer's binocular vision discovering it instead?
(and would that run the risk of damaging the viewer's eye-sight?)

 

[ihamoitc2005]

Binocular summation

I think the underlying effect is something like the term 'eye dominance'?
Here's a paper on glossiness, which suggests that for 'very glossy' materials we use binocular vision, I believe there are others on other aspects of material identification:
http://www.journalofvision.org/content/4/9/4.full

The correct term is "Binocular Summation"

This is where each eye sees different aspect of the same object and the brain takes both informations and creates an understanding. For example, camouflaged object is easier to see with stereo vision.

It is true that some people do not have proper stereo vision. Easy way to test this.

Hold a pencil in front of your face. If you see 2 pencils you have stereo vision. If you only see 1 pencil, you do not have stereo vision.

 

[JB9861]

Nice time lapse video on Red Dead. Didn't know there were several others outside of GTA 4 & crackdown. Guess I'll check them out after golf tomorrow.

http://www.eurogamer.net/articles/digitalfoundry-red-dead-redemption-time-lapse

 

[Bigduo209]

Is there enough games to do a new face-off? I'm not sure since I haven't checked any new releases.

 

[(((interference)))]

Was reading the latest face off
http://www.eurogamer.net/articles/digitalfoundry-xbox360-vs-ps3-round-27-face-off?page=1

I dont like how grandmaster downplays the 360 titles that run with 2xMSAA while the PS3 counterparts have none, this is not the first time i've noticed this either.

It's a pretty significant difference if you ask me, considering how expensive AA is.

 

[AlNom]

Was reading the latest face off
http://www.eurogamer.net/articles/digitalfoundry-xbox360-vs-ps3-round-27-face-off?page=1

I dont like how grandmaster downplays the 360 titles that run with 2xMSAA while the PS3 counterparts have none, this is not the first time i've noticed this either.

It's a pretty significant difference if you ask me, considering how expensive AA is.

Well, if you look at the darkness of say Transformers, and the fact that not all edges are properly AA'd, I'd be hardpressed to see how that is "significant". In that game, the specular aliasing is much more offensive and MSAA doesn't fix that here anyway. In their case, 2xAA shouldn't be too expensive as they seem to only use it on the passes before lighting or alpha. Same for singularity's dark levels . Aliasing is just not that big a factor. If it's just about bullet point features, you're not really understanding the point of the face-offs. The framerate analysis is probably the more important in these two cases.

 

[BadTB25]

Well, if you look at the darkness of say Transformers, and the fact that not all edges are properly AA'd, I'd be hardpressed to see how that is "significant". In that game, the specular aliasing is much more offensive and MSAA doesn't fix that here anyway. In their case, 2xAA shouldn't be too expensive as they seem to only use it on the passes before lighting or alpha. Same for singularity's dark levels . Aliasing is just not that big a factor. If it's just about bullet point features, you're not really understanding the point of the face-offs. The framerate analysis is probably the more important in these two cases.

The tearing and resolution is also very important to me.

 

[(((interference)))]

Well, if you look at the darkness of say Transformers, and the fact that not all edges are properly AA'd, I'd be hardpressed to see how that is "significant". In that game, the specular aliasing is much more offensive and MSAA doesn't fix that here anyway. In their case, 2xAA shouldn't be too expensive as they seem to only use it on the passes before lighting or alpha. Same for singularity's dark levels . Aliasing is just not that big a factor. If it's just about bullet point features, you're not really understanding the point of the face-offs. The framerate analysis is probably the more important in these two cases.

But they both are sub HD as well, sub HD with no AA generally looks horrible, see Halo 3.

And in any case he makes much of small differences in framerate, which to most people would be unnoticeable, so why play down AA vs no AA?

 

[assurdum]

Edit

 

[AlNom]

But they both are sub HD as well, sub HD with no AA generally looks horrible, see Halo 3.

The problem is the MSAA results are incorrect most of the time due to the way they do lighting and alpha passes. If you lose the MSAA result when anything is lit, how is it a significant win (It just looks like no AA)