1. Introduction. Terminology. Calculating IOD’s.

 

Why even say this? Well lots of people work in 3D already – Maya, 3ds-max and others…. sorry if I left you out – but you get the idea, you can build 3D models and animate them with lights and cameras but deliver it to a single screen. These are just the regular effects we have been doing for ages.

What we’re talking about here is Stereo Vision – separate images for the left and right eyes that create a stereo 3D illusion.

So lets call this Stereo.

You can still do 3D effects – in fact they can look great! To turn them into stereo 3D you render twice – once for each eye - and IODcalc will help you find the correct offset between those cameras, just like it does for live action photography. There are many considerations here – such as building models for your backgrounds where you used to get away with flat matt paintings – they will look just like that – flat paintings – if you don’t at least wrap them onto a surface.

OK, so the principle is very simple – you photograph the scene with two cameras, one representing the left and the other the right eye of the viewer.

There are many ways to deliver stereo footage to your audience. Passive polarisation (eg RealD), colour band (eg Dolby), anaglyph (eg red/cyan, green/magenta, blue/amber), shutter glasses (eg expand). What all these systems do is deliver the image you shot with your left camera into the audiences left eye and the right camera images to the right eye.

You can deliver your footage using any of these methods from the stereo capture – and in fact by dropping one eye you can revert to mono 2D and deliver that asset to conventional displays as well. That’s a discussion for post and deliverable – we’ll get to that – for now we’re going to deal with capturing the material correctly – on location or in pixel land if you’re working in a 3D package.

OK before you get to the maths! – what is this doing to my pictures?
- IOD (inter-ocular distance) controls the amount of perceived depth in the scene. Convergence controls where the screen plane is; ie what sticks out of or into the screen.

IOD controls depth
IOD drives perceived depth, and you can use the roundness factor that IODcalc produces for you to get an idea of where you are.

Imagine a scene photographed using a standard lens with an IOD of 65mm. This would simulate Human vision – a natural depth and roundness to objects in the scene. We call that a “roundness factor” of 1.0.

As you increase IOD the perception of depth increases and objects stretch – soccer balls start to look more like footballs oriented in/out of the screen.

Objects also start to look small. It’s sometimes called the “dollhouse effect.” What’s happening is an illusion since these objects are the same actual size on screen, but large IOD’s look as if you are a giant with your eyes wide apart and everything looks smaller to you.

The reverse is also true. If you reduce the IOD you will compress the perceived depth in the scene. All the way down to zero IOD at which point the two cameras are photographing the same thing from the same point and there is no longer any stereo information – the scene looks flat like a 2D image.

NB you will definitely need a beam splitting rig to get to small or zero IOD’s.

This can be an interesting effect and you can change the IOD during a shot – to screw with people’s perception of depth – say, if that suits the narrative.

Convergence controls the screen plane
Where an object is in relation to the convergence point will determine where the audience sees it, relative to the screen when they watch back your footage.

Objects at the convergence point, will be seen at the screen plane by the audience. Objects in front of the convergence point will stick out of the screen towards the audience. Objects behind the convergence point will appear behind the screen.

In natural vision you tend to converge on the object of interest and this is usually where you focus also. You can follow this trend in stereo photography and it will appear the most natural to your audience but it is not a hard rule. You can converge at a different distance to the focal distance – it just means that the sharply focused objects will appear to recede into the screen or project out of it.

These are all just extra tools in your kit – the mathematical limits we discuss later tell you where the illusion breaks but up to that point use all the tricks you can to enhance the narrative and immerse your audience as deeply in the plot as you can.

Focus and Depth of Field
We have traditionally used focus to direct the attention of the audience and you can still use this technique in stereo if you like. You also have convergence, which if coupled with focus will definitely drive attention to those areas of the screen.

There is much debate about small or large depth of field in stereo.

A large depth of field will allow your audience to look around in this fantastic 3D volume you are presenting to them, and this can be immersive and beautiful.

A shallow depth of field will generally guide them to look at the objects that are sharp. Just as in mono single camera photography there is little detail to be gained from out of focus areas of the screen.

Some people find out of focus backgrounds in stereo more difficult than in mono… you might be one of them!

Mathematically there is no reason you cannot do it – treat it subjectively.

It is not an excuse to exceed comfortable divergence. Just because an area of the frame is soft does not mean that you can force a whole lot more

IOD into the scene and try to “enhance the 3D” effect – you’ll make it pretty unwatchable if you do.

See the section below on “depth of coherence” which is the stereo analogue to “depth of field”.

The nature of the illusion
I called it an illusion before. This is quite correct – there is no 3D volume hanging in the screen but we trick the eyes into seeing it.
There are some limits you need to stay within to maintain the illusion – if you do this you will produce an immersive visceral experience for your audience.

Not to say be boring – far from it – we want to maintain the illusion so the experience is stronger.

If you look at 3D footage with the glasses off you see two similar images offset horizontally to each other. This is the parallax difference from the cameras being in slightly different positions when the pictures were shot – just like your eyes are offset horizontally.

We need to make sure that these offsets in the image do not exceed what the audience can accommodate otherwise the illusion will break.

There are two limits:

Background Divergence
When you look at a distant object, your eyes both look straight ahead. If you drew lines from your eyes to that object they would be pretty much parallel. If you think about this in terms of presenting those images to your eyes on a screen, they are exactly as far apart as your eyes are.

The average Human eye spacing is around 65mm (~2.5”) – give or take – so you never want to present any background divergence greater than this or you will be asking the audiences’ eyes to bend Outwards to accommodate it – and they HATE doing that!

We mean actual real world dimensions on the screen your audience is watching, so it will depend on the size of the screen the footage is displayed on. For TV and small projections it’s not generally a problem but if your footage will be seen in the cinema you need to be very careful. IODcalc will take care of all this.

 

Overall Divergence
Watching the illusion is not really the same as being there. (oh really?) What I mean is that when you look at something in the real world you normally focus on it and converge on it and then you pretty much ignore everything else.

You don’t have that ability in the cinema because everything is presented on the screen at the same distance from the viewer so we have to make sure that no matter where the audience looks on the screen, they are able to accommodate the offset in the images easily and quickly. If they cannot you will at best produce eye/brain strain and at worst break the illusion.

It is generally agreed that about 3% total offset within the image is a safe limit. However, you should experiment with this based on the content, the size of the screen and environment it will be exhibited in. Fast tight action vs. slow sweeping landscapes…very different accommodation limits.
These are the sorts of decisions that a stereographer will help you make.

It’s not just mathematical limits – even though exceeding them is a sure fire way to make your material unwatchable!

 

Terminology
There are a few new parameters to deal with and some new terminology surrounding Stereo footage.

The new parameters are the Inter-Ocular Distance (IOD) and the convergence distance.

The IOD is the distance between the axis of the two lenses, and the convergence distance is the distance from the camera to the object they are both pointing at:

Some people also refer to the convergence angle – which is the angle between these two axes – zero being parallel. It is often easier and more accurate to measure the distance than the angle in practice but they represent the same thing for any given IOD.

 

Effect of IOD
Changing the IOD affects the horizontal offset for all objects not exactly at the convergence point.

The convergence point is where the cameras are both pointing so the two images of the object at this point, from the left & right cameras, will appear on top of each other on the screen.

The two images are still subtly different producing a stereo image but the horizontal offset between these images is zero at this point.

At all other distances from the cameras there is a horizontal offset between these left & right images. By increasing the IOD, we make both the foreground offsets (convergence) and the background offsets (divergence), greater. Conversely, by reducing the IOD you can see that all these offsets get smaller.

IOD is the most important parameter you have for controlling the stereo effect of your footage.

Increasing the IOD will increase the perceived depth in the scene. To enhance the stereo effect you will usually aim to increase the IOD.

Eventually you reach either the background divergence or the overall divergence limit.

It all depends on the placement of objects in the scene in terms of their distance from the camera and where you want the convergence point to be.

What we need to achieve is a comfortable stereo image that the audience can easily resolve…. at least covering the range of distance in the scene.

We call this the “depth of coherence.”

 

Depth of Coherence
The range of distances from the camera over which the subjects will not exceed divergence limits. This range changes primarily with IOD.

To understand this better we can use an analogy to depth of field.

As you reduce the size of the iris you increase the depth of field – that range of distances from the camera over which subjects will appear sharp.

As you reduce the IOD you increase the depth of coherence – that range of distances from the camera over which the subjects will not exceed divergence limits.

Shoot Converged or Parallel
You don’t have to converge the cameras when you shoot – you can just leave them parallel and then slide the images left/right past each other in post – effectively a post convergence.

There are arguments either way for this and the answer is that there is no wrong answer!

It is dependent on the situation as there are advantages and disadvantages to both methods.

For example shooting parallel is quicker because you don’t have to adjust your camera shot to shot, but it does mean that you will need to zoom into the image more in post to cover the edges when you post converge.

There are discussions about which method deals better with lens distortion – a mute point of course if your film is animated. The more lens distortion there is, the more you need to warp the picture in post to align the left right images vertically….and this costs resolution.

This should be a discussion with your stereographer and the answer might vary production to production or even shot to shot.

Even if you shoot parallel and converge in post, it is still very important to tell IODcalc where you want the convergence point to be, as this still affects the calculations the same way.

Be sure to use the “Zoom contingency” parameter, so that IODcalc will allow for the degree of blow up you intend in post – otherwise you might find your divergences are too large after the blow up and you’ll wish you shot with a smaller IOD.

How much IOD is enough?
Up to the mathematical limit, you can use as much IOD as you subjectively want to find depth in the scene. Take a look at the Roundness factor, whichIODcalc computes for you to see how the perceived depth in the shot compares to the shots around it in the script. You can use this parameter as part of the storytelling narrative – either matching the roundness for continuity reasons or changing it up or down to change the mood of the scene.

So what is this mathematical limit?
There are two which we mentioned briefly before – the background divergence and the total divergence.
Let’s run a real world example in IODcalc to give you a feel for what’s happening.

Example – background divergence limiting

Try setting IODcalc up as follows:   These are the defaults so you can just hit Reset to Defaults if you like.

This is a metric example – sorry for those of you working in feet but the maths and principles are the same – stay with me !

Setup the main screen as:   If your target roundness is not 1.0 reset this by holding the roundness button till it flashes. You’ll see that the roundness display is actually 0.96 however… Even though IODcalc is aiming for a roundness factor of 1.0 this is not possible given the limits and distances chosen.

Why?
 

Look at the screen offset data panel…

The foreground offset is 40mm or 0.4% – no problem there.

The background offset is 65mm or 0.65% – ahh there it is – we have hit the background divergence limit and IODcalc is stopping us from bending our audiences eyes outwards.

You can see IODcalc is doing its best in terms of roundness because the USE IOD and MAX IOD are the same – in other words IODcalc wants you to use right up to the maximum mathematically permissible IOD – and even so you are not quite getting a roundness of 1.0

This does not really matter – you will still be producing a stereo image – there is no need to always work at a roundness of 1. That information is there to help you maintain continuity or transitions in mood and to let you know what you are doing.

Try changing just the convergence point distance:   If you make it smaller you will see that IODcalc further reduces the IOD.

This is because if you move the convergence point closer to the cameras without changing the IOD, the backgrounds would become further offset.

Example – foreground convergence limiting

Now try making the Convergence Point further away – say 150m.   You can see that the background divergence has reduced and IODcalc happily suggests a USE IOD of 53.3mm to achieve a roundness of 1.0 If you’re seeing some other numbers there try resetting the roundness target to 1.0. You can just hold the roundness button down for a sec till it flashes.

In fact IODcalc would be happy for you to use an IOD of 145.1mm without exceeding the mathematical limits.

What would happen if we did this?   Increase your roundness target by tapping or holding the up button next to it and watch the roundness factor and USE IOD increase. As you do this you will notice that the screen offsets are also changing – you are increasing the IOD and the divergences in the image are increasing commensurately.

Not so much in the background because our convergence point is quite near the background distance and the driving parameter is now the foreground distance.

In fact if you push your USE IOD right up to max, ie 145.1mm at a roundness of 2.72, you will see that we have hit the total divergence limit of 3% with a background divergence of just 4mm or 0.04%.

Have a bit more of a play with it to get used to the effect of changing these distances on the divergence limits – which will hit first, background or foreground.

After that try changing the focal length. This, along with sensor size, are used by IODcalc to determine the field of view (FoV) and this directly affects all these offsets.

Shorter focal lengths means a wider field of view and you can always use more IOD in any given scene if you go to shorter lenses.

The reverse is also true – on longer lenses you will need to reduce your IOD.

  

…and cut
Go out and play with the rig and use IODcalc – just in the test bay, or anywhere.

Try and set up live stereo monitoring – it doesn’t matter what type, passive polar will give you full color but even anaglyph will let you see the 3D effect.

Get your hands on the wheels and test what we have been talking about – find those limits, observe the effects. These are fantastic new tools in your kit to enhance the story telling process.

You can download IODcalc for your iPhone or iPod Touch from the App store.

For rigs and services or if you have a question or suggestion please contact us at IODcalc@e3dia.com

Enjoy!