To toe in or not to toe in, that's the question

[Sareesh Sudhakaran]

Originally Posted by Alister Chapman

You should get perfectly acceptable shots either parallel or toe'd in. Provided the camera are truly parallel then that's the safer way to shoot if you are at all unsure about your abilities or rig setup. Toe in gives you more control over on screen depth and usually give a slightly rounder image which looks more natural, but needs more care and good on set monitoring.

The longer the focal length of the lens the more obvious any rig alignment errors will be. For most S3D shoots I will try to keep to a focal length of less than 75mm (on a 35mm sensor). Typically lenses that come close to the human field of view make shooting good S3D easier. Very wide lenses will be problematic with mirror rigs as you will often end up getting the extremes of the mirror or second cameras in the shot and with side by side rigs it can hard to get the cameras close enough together. A very wide lens will give you very flat looking 3D, so you end up moving the cameras apart to add depth, which then leads to cardboarding. 35 to 50mm is an easy range to work with.

I'm helping to run a 3D event in Chennai at the end of the month.

Thank you Alister...that was the answer I was looking for. Can you tell me more about the Chennai event? Would love to know more. Thanks.

[Leonard Levy]

In 3D movie making by Bernard Mendiburu as I recall he suggests never using a lens longer than 30mm, but he doesn't say what format that refers to. I took that to mean its the lens angle that matters not the format.

This implies you'd get a wider range of lens choices on a smaller format. i.e 2/3" video would be easier to shoot 3D than 35mm and a 1/2" chip or 1/3" chip even easier.

His advice sounds way more conservative than what Alister is recommending ( 35 - 50) but that book is woefully short on addressing these kinds of issues and maybe he did have a format in mind. Way better info here.

lenny levy

[Alister Chapman]

Originally Posted by Mathew Orman

"Corrective lenses will still not eliminate cardboarding"

Wrong again, corrective lenses make distant object blend with background
totally flat as perceived by human eyes. So there is no cardboarding effect observed.

Well yes of course, if you make the image small enough that you can no longer resolve the disparity in any of the image then yes it will be flat, without cardboarding, but it will also be tiny. But one thing your forgetting is that finite camera resolution limits depth resolution. As the cameras would have been unable to resolve the depth in the players themselves so they would still be flat planes until eventually you corrected the image enough to make the entire image flat. It's arguable that cardboarding can occur in 2D images as well as 3D images. Long shots done with tracking cameras or camera jibs can appear to break down into planes as the camera moves because or brain is confused by the depth cues it is seeing.

This discussion is about how S3D is currently made and viewed, it's not about hypothetical theories about perfect stereoscopic vision replication, which seems to ignor the way our brain works using combinations of depth cues, not just matching focal lengths and viewing distances.

Assuming the viewer is perpendicular to the screen and has eyes 65mm apart, then 65mm I don't need to calculate that.

[Leonard Levy]

Mathew,

Maybe I'm dense but i don't understand what orthographic stereo is, and it sounds like your describing a type of cinema that doesn't exist in most of our practical experience. Does what your talking about have practical application?

Also I hope we can keep this discussion respectful and courteous. Its been the best discussion of the subject I've seen anywhere so far and I'd hate for it t get sidetracked.

[Alister Chapman]

Leonard. Sensor size makes no difference it is the field of view that matters because when you deviate greatly from a natural FoV it's much harder to produce realistic looking images.

At the moment no one is really sure exactly what is acceptable or desirable. Do viewers want a natural image or do they want forced depth. For a narrative piece you possibly can stay very natural, but sports in particular present all kinds of problems some of which come about because of the limited resolution of the cameras and viewing systems. I would not want to watch a soccer match all shot with wide shots on TV. Longer shots do take you closer to the action but then you have cardboarding and other undesireable effects. In 2D we just accept foreshortening as normal but in 3D it is more objectionable.... or is it? Do we accept it and live with it because forced depth helps tell you where the ball is going in 3D space. These are questions that are still un answered.

[Alister Chapman]

Othographic reproduction is the 1 to 1 reproduction of something. So ortho stereo would have cameras that have human FoV toed in on the most distant subject in the screen. The footage is then shown on a screen that matches the human FoV with everything in negative parallax so what you see is an entirely natural view.

The problem is that for true ortho stereo it requires that the screen is as far away from the viewer as the most distant object in the scene as everything is in negative parallax and thus in front of the screen. That's fine for a fixed scene of say a room, where you could match the viewer to screen distance to the back wall of the room. But when you are looking at exteriors with massive depth that presents a problem as the screen would need to be miles away and massive. So in practice you place infinity (or your back wall) behind the screen so now you have both negative and positive parallax but now it's no longer true otho stereo as to do this requires some kind of toe in via the projectors or electronically which can then introduce some of the issues we are all familiar with. True othos stereo also only has one single viewing position, the one that matches the camera so only the one person sat in that spot will see an ortho stereo image, everyone else see's something not quite ortho stereo.

The nearest I've seen to ortho stereo is some of the early IMAX 3D films where they tried to match camera FoV to Human FoV. In the early days this was pretty close as all IMAX screens were the same size and the stadium seating means there is not so much variation from seat to seat in screen to viewer distance. These days it's not so good.

[Mathew Orman]

Originally Posted by Alister Chapman

Well yes of course, if you make the image small enough that you can no longer resolve the disparity in any of the image then yes it will be flat, without cardboarding, but it will also be tiny. But one thing your forgetting is that finite camera resolution limits depth resolution. As the cameras would have been unable to resolve the depth in the players themselves so they would still be flat planes until eventually you corrected the image enough to make the entire image flat

This discussion is about how S3D is currently made and viewed, it's not about hypothetical theories about perfect stereoscopic vision replication, which seems to ignor the way our brain works using combinations of depth cues, not just matching focal lengths and viewing distances.

Assuming the viewer is perpendicular to the screen and has eyes 65mm apart, then 65mm I don't need to calculate that.

Yes,
and currently S3D is totally distorted for the sake of parallax size on the big screen since if it is OK on big screen it will be miniaturized on all small ones and parallax will be always less than infinity for the backgrounds.

This type of fallacy and disinformation of public is now being rejected and exposed. See example:

3D, Who Needs It? | John C. Dvorak | PCMag.com

Also your calculation shows lack of understanding of real geometry of human stereoscopic vision.
The infinity parallax for screen that is 30 feet away is: 860 mm or 34 inch

Mathew Orman

[Alister Chapman]

Originally Posted by Mathew Orman

Also your calculation shows lack of understanding of real geometry of human stereoscopic vision.
The infinity parallax for screen that is 30 feet away is: 860 mm or 34 inch

Mathew Orman

And exactly how did you come to that figure? How far from the screen is the viewer seated? Our eyes do not diverge when viewing infinity.

Perhaps I don't understand the true geometry of human vision, but I do know what can and can't be done in an S3D production and what does or does not look good to the vast majority of people and this is what is important. Film making is a trick, it is a gimmick. We use actors because thay are good at exagerating certain phrases or actions to help enhance the story. Take explosions in films, they are normally full of fire and flame when a real bomb tends to be a shock wave with little fire. But which does the viewer enjoy watching, big, dramatic fireballs or shockwaves?

Are you seriously telling me that I and just about every other person making S3D films is wrong and that you are the only person that is correct. You do seem to forget that we are not creating 3D images but pairs of 2D images and creating an optical illusion that does not just rely on human vision geometry but also uses other depth cues, in particular scale. It's very easy to trick people into thinking they are looking at objects that are infinity simply using scale or even removing any other depth cues. A well set up planetarium for example will have people believing that they are looking at stars at infinite distance even though the projected 2D dot is only 30ft away. Our eyes do not measure distance they estimate depth ratios and relationships our brain then compare how far A is from B and then using scale judges depth. That's why golfers find it hard to judge the distance to a green unless there is a flag to help with scale. Put a different size flag in the hole and see what happens.

We are generating an optical illusion, not a mathematically correct 3 dimensional image and like all illusions there is more to it than pure science.

[Giroud Francois]

Yes, Mr Dvroak is right about DOF, but unfortunately DOF is the Hype today with DSLR.
Historically DOF had some reasons to exist, but not today (except for some artistic will).
It is really hard to match DOF and S3D to give a watchable picture. I think Avatar was not so bad at it.
Unfortunately, as soon as you are shooting real world, you do not get the same possibilities than getting pure computer rendered scenes.

[Mathew Orman]

Originally Posted by Alister Chapman

Othographic reproduction is the 1 to 1 reproduction of something. So ortho stereo would have cameras that have human FoV toed in on the most distant subject in the screen. The footage is then shown on a screen that matches the human FoV with everything in negative parallax so what you see is an entirely natural view.

The problem is that for true ortho stereo it requires that the screen is as far away from the viewer as the most distant object in the scene as everything is in negative parallax and thus in front of the screen. That's fine for a fixed scene of say a room, where you could match the viewer to screen distance to the back wall of the room. But when you are looking at exteriors with massive depth that presents a problem as the screen would need to be miles away and massive. So in practice you place infinity (or your back wall) behind the screen so now you have both negative and positive parallax but now it's no longer true otho stereo as to do this requires some kind of toe in via the projectors or electronically which can then introduce some of the issues we are all familiar with. True othos stereo also only has one single viewing position, the one that matches the camera so only the one person sat in that spot will see an ortho stereo image, everyone else see's something not quite ortho stereo.

The nearest I've seen to ortho stereo is some of the early IMAX 3D films where they tried to match camera FoV to Human FoV. In the early days this was pretty close as all IMAX screens were the same size and the stadium seating means there is not so much variation from seat to seat in screen to viewer distance. These days it's not so good.

Wrong again,
the requirements are that camera geometry matches the viewer geometry and in such stereo window size that is defined by camera geometry is the same as the projection screen and both FOVs are the same.
The toe-in after keystone correction is no different than off-axis camera.
What is in front or behind the screen is only defined by stereo window position within a scene.
Finally both toe-in and off-axis camera geometry do not match real geometry of human stereoscopic vision.
And it looks like none of the stereographers can define the correct stereoscopic camera geometry to match human vision.
That is why when one ask to define infinity parallax they always quote the human eye distance.

No orthographic stereo view example yet?

Mathew Orman

[Mathew Orman]

Quote:
Originally Posted by Mathew Orman View Post
Also your calculation shows lack of understanding of real geometry of human stereoscopic vision.
The infinity parallax for screen that is 30 feet away is: 860 mm or 34 inch

Mathew Orman

And exactly how did you come to that figure? How far from the screen is the viewer seated? Our eyes do not diverge when viewing infinity.

// how far? it states above: 30 feet.
// yes they do not diverge but 65 mm vanishes to 0 or single point at infinity
// and parallax as perceived by human eyes is an angular size not linear distance. And that is why for 30 feet distance it translates to 860 mm.



Perhaps I don't understand the true geometry of human vision, but I do know what can and can't be done in an S3D production and what does or does not look good to the vast majority of people and this is what is important. Film making is a trick, it is a gimmick. We use actors because thay are good at exagerating certain phrases or actions to help enhance the story. Take explosions in films, they are normally full of fire and flame when a real bomb tends to be a shock wave with little fire. But which does the viewer enjoy watching, big, dramatic fireballs or shockwaves?

// yes, no one is restricting fantasy but distorting geometry creates confusion, eye strain and unpleasant experience and as some point it out the distortion is made as a compromise for the single viewer position requirement of real undistorted geometry. Zoom is the worst cheep trick that is abused for the excuse of story telling and the fact that the narrowest DOF is obtained with long focal length.

Are you seriously telling me that I and just about every other person making S3D films is wrong and that you are the only person that is correct.

// No, there are few people still alive that have the full understanding.

You do seem to forget that we are not creating 3D images but pairs of 2D images and creating an optical illusion that does not just rely on human vision geometry but also uses other depth cues, in particular scale. It's very easy to trick people into thinking they are looking at objects that are infinity simply using scale or even removing any other depth cues.

// Wrong again, simulated or real stereoscopic image is no different if we
gaze at the center. Our brain senses and distinguishes distance based on triangulation and parallax is angular size and that is why the rule requires perceptive to be undistorted. Human stereoscopic vision is a precise 3D navigation apparatus that estimates spacial distribution of object with extreme accuracy and serve the porpoise of collision free commuting well.
Also the vision is cyclopean and we do not see split images of left and right eye. We might see split objects if for example we look at horizon the finger in front would split in two.
//

A well set up planetarium for example will have people believing that they are looking at stars at infinite distance even though the projected 2D dot is only 30ft away. Our eyes do not measure distance they estimate depth ratios and relationships our brain then compare how far A is from B and then using scale judges depth. That's why golfers find it hard to judge the distance to a green unless there is a flag to help with scale. Put a different size flag in the hole and see what happens.

// Wrong again,
mirror stereoscope is required on such planetarium screen to see 2D image at infinity.
What they see in a planetarium are light spots on the flat screen 30 feet away.
Human vision is capable of resolving distances from 1 inch to 1000 feet
which is only limited by numbers of view cones that form retina.
//

We are generating an optical illusion, not a mathematically correct 3 dimensional image and like all illusions there is more to it than pure science.

// Wrong again,
stereoscopic vision is a precise instrument and optical illusions are not observed in stereo. But if you know of such then please provide an example
because I have never see an optical illusion in stereo.
You can trick human brain with single eye image but not in stereo.
//
__________________
Alister Chapman, Film-Maker/Stormchaser XDCAM-USER.com My XDCAM site and blog. Alister Chapman DoP, Stereographer my home page.
Alister Chapman is online now Report Post Reply With Quote


// Mathew Orman

[Alister Chapman]

Originally Posted by Mathew Orman

Wrong again,
the requirements are that camera geometry matches the viewer geometry and in such stereo window size that is defined by camera geometry is the same as the projection screen and both FOVs are the same.
The toe-in after keystone correction is no different than off-axis camera.

Isn't that what I described, only perhaps not so elegantly?

Originally Posted by Mathew Orman

Finally both toe-in and off-axis camera geometry do not match real geometry of human stereoscopic vision.

Oh really. My eyes most certainly toe in, just take a look at anyone that looks at the end of their nose or at their own hands. Their eyes toe in, it's plain to see.

Originally Posted by Mathew Orman

That is why when one ask to define infinity parallax they always quote the human eye distance.

So if i were to take a couple of infinitely long pieces of string and stretch them from my eyes to an infinitely distant object would those two strings not in essence be parallel and would my eyes not be looking down along those strings? The assumption is that when viewing an infinitely distant object that our eye's will be parallel. If I then sit in the cinema and stretch two parallel pieces of string from my eye's to the screen the ends of the string would be 65mm apart on the screen and as above my eyes would be looking along those strings. If I used your calculated 34" separation those string would most certainly be diverging. Logic says that to view to images separated by 34" my eyes would have to be looking very slightly apart as they follow those strings to the screen, this is not natural and not the same as looking at an object truly at infinity. Experience tells us that when a viewers eyes are forced to diverge by excessive amounts it creates problems such as eyestrain or headaches

So please enlighten us all Mathew as to how we should be doing it. So far I have yet to see a single constructive post from you on these boards. All your post have ever been is negative posts telling us that everyone except you is wrong. Please instead of just telling us that we are all wrong tell us what we should be doing and not theoretically but in real world practical terms.

[Mathew Orman]

Originally Posted by Leonard Levy

Mathew,

Maybe I'm dense but i don't understand what orthographic stereo is, and it sounds like your describing a type of cinema that doesn't exist in most of our practical experience. Does what your talking about have practical application?

Also I hope we can keep this discussion respectful and courteous. Its been the best discussion of the subject I've seen anywhere so far and I'd hate for it t get sidetracked.

Google for orthographic camera first then to create stereoscopic pair: toe-in orthographic cameras on some volume that encloses some geometrical objects.
It can be easily done using CG packages such as Maya, Lightwave, etc.

Here is some info for the new stereoscopic cinema system:

3D Cinematography


Mathew Orman

[Mathew Orman]

Alister Chapman;1556435]Isn't that what I described, only perhaps not so elegantly?

// No, you've quoted "would have cameras that have human FoV toed in on the most distant subject in the screen"
which is wrong.


Oh really. My eyes most certainly toe in, just take a look at anyone that looks at the end of their nose or at their own hands. Their eyes toe in, it's plain to see.

// I guess you do not know how to setup off-axis camera that has a small stereo window in-front of your nose.



So if i were to take a couple of infinitely long pieces of string and stretch them from my eyes to an infinitely distant object would those two strings not in essence be parallel and would my eyes not be looking down along those strings?

// no we do not envision two images the vision is cyclopean.
also do not need the strings to infinity just 30 feet would prove that you see two lines merging to single point at the other end. Or just stand in-between rails on rail road track and see if the rails are parallel.
On the other hand you could position string such that the front and the end
occupy the same FOV and you immediately notice parallelism.
//



ortographic proiection

The assumption is that when viewing an infinitely distant object that our eye's will be parallel. If I then sit in the cinema and stretch two parallel pieces of string from my eye's to the screen the ends of the string would be 65mm apart on the screen and as above my eyes would be looking along those strings. If I used your calculated 34" separation those string would most certainly be diverging. Logic says that to view to images separated by 34" my eyes would have to be looking very slightly apart as they follow those strings to the screen, this is not natural and not the same as looking at an object truly at infinity. Experience tells us that when a viewers eyes are forced to diverge by excessive amounts it creates problems such as eyestrain or headaches

// wrong again,
did you see the 30 feet infinity parallax example and had problem with eye strain?
do you know what is your desktop monitor infinity parallax for your favorite view angle or screen distance?
//

So please enlighten us all Mathew as to how we should be doing it. So far I have yet to see a single constructive post from you on these boards. All your post have ever been is negative posts telling us that everyone except you is wrong. Please instead of just telling us that we are all wrong tell us what we should be doing and not theoretically but in real world practical terms.

//No, that is you impression.
Everyone else is just having a problem on how to distort the geometry and at the same time make it look good.
All I am trying to convey is the fact that real stereoscopic geometry is simple and extremely easy to implement yielding
real and undistorted immersion in original scene's volume.
But I guess no one on this forum has experience such immersion or is willing to create one.
//


//Finally to create what I am on about one should:
Make an off-axis camera rig with 65 mm base and stereo window set for the target screen size
and just shoot using dolly instead of zoom and if objects intersects stereo-window's sides use stereoscopic masking in post.
The content will play perfect on target screen that matches the stereo window and for other screen sizes use floating stereo windows and you will get the same undistorted immersion.

// Mathew Orman

[Tony Asch]

Rather than continuing a somewhat abstract theoretical discussion, perhaps Alister and Mathew would share some real world examples, in the form of 3D stills or video/film, which demonstrate the points they are making, and more importantly allow the rest of us, including some people who are not stereoscopic-literate, to evaluate the aesthetic qualities of Mathew's and Alister's approaches.

After all, the final arbiter should be the audiences of 3D productions. Do the production techniques enhance the perception of 3D? Do they provide a comfortable viewing experience? Do they offer added value to the story-telling art?

To the extent that these demos require a specialized viewing system/environment, I'm sure Alister and Mathew can direct us to such locations. I look forward to the experience!