Game Engine Gems 2

 163

11

Making3DStereoscopicGames

Sébastien Schertenleib

Sony Computer Entertainment Europe

11.1Introduction

With the large variety of 3D content being made available (sports events, movies,

TV, photos, games, etc.), stereoscopic 3D is gaining momentum. With the sup-

port for 3D content on the PC and game consoles such as the PlayStation 3 and

Nintendo 3DS, it is likely that it will become even more widespread. In this chap-

ter, we present some topics that need to be considered when creating or convert-

ing a game to stereoscopic 3D. We also present some optimization techniques

that are targeted to improving both the run-time performance and visual fidelity.

11.2HowStereoscopic3DWorks

Stereoscopic 3D is produced by creating two separate images (one for each eye)

that are then displayed on a 3D screen, as shown in Figure 11.1. Depending on

the technology in place, those two images are then separated for the correct eyes

through some means. The three major approaches are the following:

■ Active shutter glasses. The screen alternately displays the left and right im-

ages and sends a signal to the LCD screen in the lens for each eye, blocking

or transmitting the view as necessary.

■ Passive polarized glasses. The screen is paired with adjacent right and left

images using orthogonal polarizations. The filter on each eye blocks the or-

thogonally polarized light, allowing each eye to see only the intended image.

■ Parallax barrier. The screen features a layer of material with some slits

placed in front of it, allowing each eye to see a different set of pixels without

glasses, but with restricted view angles.

164 11.Making3DStereoscopicGames

Figure 11.1. Creating a stereoscopic 3D scene.

11.3HowtoSetUptheVirtual3DCameras

Contrary to other media, video games have the luxury of being able to control the

camera properties directly. As we mentioned earlier, stereoscopic 3D requires

that we set up two distinct cameras. One possible solution is illustrated in Figure

11.2(a), where we use a simple offset to move the left and right cameras. This

approach results in a large portion of each image being visible to only one eye, as

shown by the arrows. This tends to generate strong eye strain. Therefore, an al-

ternative approach that is sometimes used with a stereoscopic camcorder is to toe

in both cameras by rotating them inward, as shown in Figure 11.2(b). However,

the convergence is no longer parallel to the screen, producing a vertical parallax

deviation when the camera rotates upward or downward. This is unnatural and

uncomfortable for the user. To circumvent these shortcomings, the scheme de-

picted in Figure 11.2(c) consists of using parallel cameras with an asymmetric

projection that minimizes the zone covered by a single image while avoiding ver-

tical parallax. This provides a much more comfortable experience.

Glasses

11.3HowtoSetUptheVirtual3DCameras 165

Figure 11.2. (a) A simple offset is applied to the left and right cameras. (b) Both cameras

are rotated inward. (c) The cameras have parallel view directions by use asymmetric pro-

jections. Configurations (a) and (b) lead to issues that deteriorate the stereoscopic 3D

experience. Configuration (c) avoids those shortcomings by using asymmetric projection

matrices.

With this model, the usual projection matrix

p

roj

M

given by

proj

2

00

2

00

2

00

00 1 0

nrl

rl rl

ntb

tb tb

M

nf nf

nf nf



























































changes because it is no longer the case that

0rl



 and 0

t

b



 . The off-center

view frustum that we must use is shown in Figure 11.3.

LR

LR

LR

(a) (b)

(c)