STEREOSCOPIC VISION IN DESKTOP AUGMENTED REALITY

User Performance in the Presence of Conﬂicting Depth Cues

Gustavo Rovelo

, Francisco Abad

, M.-Carmen Juan

and Emilio Camahort

Departamento de Sistemas Inform

aticos y Computaci

on, Universitat Polit

ecnica de Val

encia, Camino de Vera S/N 46022

Valencia, Spain

Instituto Universitario de Autom

atica e Inform

atica Industrial, Universitat Polit

ecnica de Val

encia, Camino de Vera S/N

46022 Valencia, Spain

Keywords:

Augmented Reality, Stereoscopic Vision, User Performance Assessment, Perceptual Issues.

Abstract:

The use of stereoscopy as a depth cue in current computer graphics technology is increasingly popular. On the

other hand, in a Desktop Augmented Reality (AR) environment, a web cam and a set of printed markers allow

the user to interact with an augmented view of the environment in the computer display. So, it seems a logical

step for Desktop Augmented Reality systems to take advantage of the available stereoscopic hardware to

improve the realism of the augmented scenes. The main problem of Desktop AR applications is that they will

be used in different and unprepared environments (wide variety of hardware, inconsistent and unpredictable

lighting conditions, etc.) Therefore, it is crucial to understand how adverse conditions affect user experience.

Furthermore, Desktop AR applications usually present conﬂicts in the depth cues presented to the users:

viewpoint offset and incorrect occlusions between real and virtual objects.

We use the within subjects experimental approach to evaluate user performance in an Augmented Reality

game. Our goal is to ﬁnd if stereoscopic graphics help to reduce the impact of the other depth cues conﬂicts

inherent to Desktop AR applications.

1 INTRODUCTION

Augmented Reality (AR) applications have become

more common beyond the research laboratory. Users

can now use AR systems at home while surﬁng the

web (Sony Pictures Digital Inc., 2009), using their

mobile phones (SPRX Mobile, 2010) or playing with

a video console (Sony Computer Entertainment Eu-

rope, 2009).

The use of stereoscopy as a depth cue in com-

puter graphics technology is increasingly popular too.

However, stereoscopy is not the only depth cue hu-

mans use to perceive the third dimension. Motion par-

allax, perspective, accommodation, shadows, texture

gradient and occlusion are other depth cues examples.

When implementing Desktop AR applications, it

is common to introduce conﬂicts in different depth

cues due to the limitation of the devices and the real-

time constraints. For example, when occlusion be-

tween virtual and real objects is not correctly solved,

virtual objects always appear on top of the real ob-

jects. The Figure 1 shows an example of this problem:

the user’s hand is occluded by the virtual object when

the hand should be visible, since it is closer to the

camera. These conﬂicting cues confuse the brain and

reduce the realism of the application. Another prob-

lem of AR systems, that is magniﬁed in Desktop AR

environments is the viewing angle offset problem:

real and virtual objects are shown in the display from

the viewpoint of the web cam, that is quite different

from the user’s viewpoint (Kruijff et al., 2010). In

professional AR systems, this problem is solved us-

ing see-through or video-see-through HMDs.

In this paper we study how users react to con-

ﬂicting depth cues when completing a spatial task in

a Desktop AR application. Particularly, we analyze if

using stereoscopic graphics to render virtual objects

increases the user’s performance. Our testbed is a

popular game for children, the Wire Loop Game. We

compare user performance using both monoscopic

and stereoscopic graphics.

We ran experiments with 32 volunteers of differ-

ent ages. For each volunteer and each game, we

collected ﬁve performance measurements. We then

statistically analyzed the measurements to determine

user performance. Additionally, each user answered

460

Rovelo G., Abad F., Juan M. and Camahort E..

STEREOSCOPIC VISION IN DESKTOP AUGMENTED REALITY - User Performance in the Presence of Conﬂicting Depth Cues.

DOI: 10.5220/0003852304600465

In Proceedings of the International Conference on Computer Graphics Theory and Applications (GRAPP-2012), pages 460-465

ISBN: 978-989-8565-02-0

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

Figure 1: Virtual objects are always drawn in front of real

objects. This ﬁgure shows the path for the easiest level of

the game.

a questionnaire about non-quantiﬁable subjective is-

sues about her game experience.

The outcome of the statistical analysis shows that

using stereoscopic graphics does not beneﬁt user per-

formance in a Desktop AR application. We found

that users performed better with monoscopic viewing.

We also found that users playing with stereoscopy

enabled ﬁrst had worse results the second time they

played (with monoscopic graphics) than users play-

ing the game with a monoscopic view during the ﬁrst

round. This effect is even more noticeable when the

task is more complex, as in the third (and last) level

of our game.

The rest of the paper is organized as follows. First,

in Section 2 we make a brief review of related work.

Then, Section 3 describes the features of our AR

game and system. In Section 4 we explain the goal

of our research, the experimental design of our study

and the user tests. Sections 5 and 6 describe and dis-

cuss our results. Finally, in Section 7 we present our

conclusions and directions for future work.

2 RELATED WORK

Many authors have studied the problems that com-

monly arise in AR applications. There are surveys

that discuss perception issues in AR systems like,

for example, occlusion, stereoscopy, shadows, visuo-

haptic collocation, user viewpoint tracking and visual

angle offset between real and virtual objects (Azuma

et al., 2001; Drascic and Milgram, 1996; Kruijff et al.,

2010).

There are also more speciﬁc studies that analyze

the impact of stereoscopic depth cues in AR sys-

tems (Lawson et al., 2002; Sands et al., 2004; Shidoji

et al., 2010). Other works study the conﬂict of han-

dling occlusion between real and virtual objects (Hou,

2003). Ruling et al propose a new algorithm to cal-

culate the correct occlusion in unprepared scenarios

where unmodeled real objects might hide virtual ob-

jects (Ruling et al., 2008).

Researchers have also studied how shadows can

improve user’s perception of depth and virtual ob-

jects (Sugano et al., 2003), or the minimal resolution

in a light map to render perceptually correct artiﬁcial

shadows (Nakano et al., 2008).

Other studies in user-centered evaluation focus on

how improving the visualization hardware can alle-

viate AR perception issues (Cakmakci et al., 2004;

Livingston et al., 2009; Liu et al., 2008).

All of the cited works try to understand and re-

duce perceptual issues in AR. Many times the solu-

tions proposed use specialized tracking and rendering

hardware. Other times only one or a few issues are

considered instead of analyzing the whole picture. In

our study we perform a user-centered evaluation to

analyze the perceptual issues that arise in a Desktop

AR application.

Our purpose is to evaluate whether stereoscopy

helps or hinders the user’s performance, taking into

account the conﬂicting depth cues in this type of AR

application. We quantify the user’s performance by

taking certain measurements described later. In the

following section we describe the game we used to

run the tests.

3 THE AR WIRE LOOP GAME

In the Wire Loop Game, players need psychomotor

coordination and good depth perception skills. The

goal is to move a wire loop along a wire path without

the loop touching the path. If this happens, the game

provides the user with some form of feedback. The

AR Wire Loop Game we developed deﬁnes three dif-

ferent game levels. Each level has a different number

of curves in the virtual wire path: one, three and ﬁve

for the easy, intermediate and hard level, respectively.

A Wii remote (Wiimote), the computer display and

its speakers are used as the output devices to provide

tactile, visual and aural feedback, respectively.

The virtual wire path and its wooden support are

rendered on top of a real ARToolKit marker. We place

the marker on a table in front of the system’s camera

and the stereoscopic display. The user can move and

orient the virtual ring moving and orienting the Wi-

imote and its attached marker.

Two situations may arise while playing the game.

First, the user may touch the wire with the ring. We

call this situation a collision event. Second, a cross-

ing event occurs when the user completely crosses the

wire with the ring. In this case the game draws a yel-

STEREOSCOPIC VISION IN DESKTOP AUGMENTED REALITY - User Performance in the Presence of Conflicting

Depth Cues

461

Markers

WiiMote

Webcam

Personal

Computer

NVidia 3D

Vision Kit

Figure 2: The components of our system: two printed mark-

ers, a PC with a NVIDIA Quadro 600 graphics card, a web

cam, a Wiimote, a 3D display and the NVIDIA GeForce 3D

Vision kit.

low virtual ring at that point (see Figure 1) and she

will not be able to ﬁnish the level unless she takes the

ring back at the point where the path was left in the

ﬁrst place. In both situations, visual, aural, and tactile

stimuli are used to provide feedback to the user.

The Wiimote is connected to a PC using Blue-

tooth. We use the Wiiuse library to communicate our

application with it (Michael Laforest, 2008). Stereo-

scopic graphics is achieved using an LG Flatron

W23630 monitor and the NVIDIA GeForce 3D Vi-

sion Kit (active stereoscopic glasses and a NVIDIA

Quadro 600 graphics card). Figure 2 illustrates our

setup. We used osgART (Looser et al., 2006) to build

our game. osgART is a toolkit that offers fast integra-

tion of virtual and real-world 3D objects.

Besides stereoscopy, our system implements other

depth cues like shadows. All virtual objects, that is,

the path and the ring, cast and receive shadows. We

also use occlusion between virtual objects as a depth

cue. Normally, part of the ring occludes the wire and

vice versa (see Figure 3).

Figure 3: The ring and the path cast shadows over the virtual

board and over each other. The occlusion is also solved

between virtual objects.

4 EXPERIMENTAL DESIGN

We performed a study with 32 right handed volun-

teers aged 21 to 43 (27.5 on average). All of them

volunteered their time. The design of our experiment

deﬁnes two factors with two possible levels each. The

ﬁrst factor is using stereoscopic graphics (enabled or

disabled). We use the Within Subjects experimen-

tal design, therefore all subjects played using stereo-

scopic and monoscopic graphics. The second fac-

tor of the experiment is the order used for each user

(stereoscopic visualization ﬁrst, then monoscopic; or

the other way around). All users played the three lev-

els of the game in the same order. Half of the users

started with the monoscopic version, the other half

with the stereoscopic version.

All subjects were sitted at about 50 cm from the

monitor. The ARToolkit marker to track the wire path

location was placed in the table in front of the moni-

tor. The web cam was located on top of a tripod to the

left of the user, to avoid interferences with her arms

while playing the game. The game starts placing the

loop at the farthest end of the path with respect to the

user.

Before playing, each participant watched an intro-

ductory 3.5 minutes video with instructions on how

to play the game. After that, a member of our team

gave the user some ﬁnal instructions. Then, the sub-

ject played the three levels (with stereoscopy enabled

or disabled). After that, the subject played again the

three levels, with the other conﬁguration.

The ﬁrst time a user plays, she spends one minute

practicing with the ﬁrst level of the game before run-

ning the experiments. The goal of each user is to ﬁn-

ish each level in the shortest possible time and with

the least amount of mistakes. Mistakes, as explained

earlier, occur when the user touches the wire with the

ring, and when the user crosses the wire with the ring.

Users were allowed to take a one minute break be-

tween levels. Once all three levels were completed,

we asked the users to ﬁll out a questionnaire. We also

gathered the user’s opinion about the game by doing

a brief interview with each one of them.

To perform our Within Subjects analysis, we

recorded the following measurements: Completion

time T (time required to complete a game level);

Number of collisions N

(number of times the user

touches the wire with the ring); Number of crossings

(number of times the user crosses the wire with the

ring); Average collision time T

(average time the user

takes to stop touching the wire after a collision) and

Average time outside the wire path T

(average time

the user takes to bring the ring back to the wire path,

after crossing it).

GRAPP 2012 - International Conference on Computer Graphics Theory and Applications

462

The times listed above do not include the times

spent by our system generating the feedback stim-

uli. These times are negligible compared to the above

measurements. For example, aural stimuli take 0.04

seconds on average to generate, while the Wiimote

vibration requires 0.03 seconds on average to be acti-

vated. The worst-case scenario frame rate of our sys-

tem occurs when stereoscopy and all three feedback

stimuli are enabled. In that case the frame rate is 28

fps. The best-case frame rate is 36 fps.

5 STATISTICAL ANALYSIS

In this section we present the outcome of the statisti-

cal analysis of the results of our experiments.

To study the data of our tests, we did a Multifactor

ANOVA analysis. We selected Fisher’s Least Square

Difference method to prove our hypothesis for all

ﬁve measurements recorded and for both monoscopic

and stereoscopic graphics. We applied the method at

the 95% conﬁdence level. We performed the analysis

using StatGraphics

 and assuming the following

null and alternative hypotheses:

: performance

= performance

: performance

6= performance

Our H

states that using stereoscopic graphics will

not improve user’s performance in our Desktop AR

testbed.

Table 1 presents the Multifactor ANOVA analysis

outcomes for every one of the ﬁve measurements for

the stereoscopy factor. We ﬁnd a statistically signiﬁ-

cant difference only for T

in Level 3; so we can not

reject H

Since we use the Within Subjects design, and ev-

ery user plays each game level twice, we analyzed the

relationship between both factors: experiment order

and use of stereoscopy. We did not ﬁnd any statis-

tically signiﬁcant relationship between both factors

for any of the ﬁve performance measurements on any

game level.

We detected the presence of outliers in the data.

For this reason we performed a cleaning process to

remove those samples. Despite this fact we can ob-

serve that the data dispersion in some cases is not

small enough.

5.1 Questionnaires and Interviews

The tests we ran with our volunteers allowed us to

gather subjective information too. We collected it us-

ing questionnaires and interviews. The questionnaires

included yes/no questions, open questions and Likert

scale questions (from one, strongly disagree, to seven,

strongly agree).

The users’ answers showed that more than 70%

of them knew the real version of game. 56.5% of the

users who knew the real version of the game liked our

game more than the real version. Almost all users

(84.3%) enjoyed playing it.

Even though users found the game easier with-

out stereoscopy, 65.6% of them reported that stereo-

scopic graphics let them better perceive object depth.

25% reported that they perceived depth better without

stereoscopic visualization and 9.4% reported no dif-

ference. These results were conﬁrmed using a second

question about the number of errors the users felt had

incurred while playing: 28.3% of the user said they

felt that stereoscopy had not helped them, 9.4% said

that there was no difference and 62.3% that the stere-

oscopy helped them to had less mistakes.

We also asked the users about the game. They said

that the stereoscopic version offered more informa-

tion about the depths of the ring and the path. How-

ever, they complained that it was sometimes hard to

accommodate their vision to focus on the objects, and

that they were not always able to see 3D correctly.

They also said to suffer ghosting (double images). Fi-

nally, camera and screen positions were mentioned to

cause discomfort.

Conversely, the use of shadows as a depth cue was

considered very useful. They were particularly useful

at the beginning of the game, when moving the ring

into the path for the ﬁrst time.

6 DISCUSSION

In our experiments, each user played through the three

levels of difﬁculty of the game twice, with and with-

out stereoscopy. Each time we took ﬁve measure-

ments: T , N

, N

, T

and T

. The ﬁrst three mea-

surements increase with the level of difﬁculty. This

is independent of using monoscopic or stereoscopic

graphics. This is due to the increased length of the

wire and the increased number of curves in the wire.

Regarding the two reaction times, we observe a

strange behavior for level 2. Users need more time to

correct the ring position after touching the wire than

in level 3. On the other hand, the use of monoscopic

or stereoscopic graphics does not affect on average

the times needed to correct the ring’s position. As for

user performance, we do not ﬁnd statistical evidence

to reject the null hypothesis for any measurement we

took. This means that, even though on average mono-

scopic results are better than stereoscopic results, sta-

STEREOSCOPIC VISION IN DESKTOP AUGMENTED REALITY - User Performance in the Presence of Conflicting

Depth Cues

463

Table 1: Multifactor ANOVA analysis outcomes: stereoscopy factor.

Dependant Variable Level 1 Level 2 Level 3

Completion time F(1, 58) = 0.87, p = 0.3555 F(1, 58) = 0.64, p = 0.4271 F(1, 58) = 0.91, p = 0.3446

Number of collisions F(1, 58) = 0.24, p = 0.6233 F(1, 62) = 0.87, p = 0.3561 F(1, 62) = 0.33, p = 0.5684

Number of Crossings F(1, 58) = 0.03, p = 0.8742 F(1, 58) = 0.20, p = 0.6573 F(1, 60) = 0.02, p = 0.8970

Avg. Collision time F(1, 56) = 0.45, p = 0.5044 F(1, 58) = 1.75, p = 0.1906 F(1, 60) = 0.25, p = 0.6170

Avg. Time outside the path F(1, 58) = 0.54, p = 0.4660 F(1, 60) = 2.22, p = 0.1418 F(1, 60) = 4.06, p = 0.0485

tistically using monoscopic or stereoscopic graphics

does not affect user performance while playing the

game.

Using stereoscopy in graphics applications im-

proves depth perception, but our experiments show

that under conﬂicting depth cues, stereoscopic graph-

ics does not improve user performance in a position

and orientation task like the one we evaluated. One

reason is the fact that the user sees the virtual objects

(and the real scene) on the display from the point of

view of the camera, not from her own point of view

(see (Azuma et al., 2001) for a detailed description

of this problem). Another reason is that occlusion be-

tween real and virtual objects is not correctly handled.

The user’s hand and the Wiimote always appear be-

hind the virtual objects. These contradictory depth

cues confuse some users. They reported better depth

perception, but the scene looked weird to them. Users

had problems accommodating their vision to the 3D

objects. Finally, some users reported eye strain after

playing the game with stereoscopic glasses. All these

perception issues are common to other AR applica-

tions, and have also been reported by (Kruijff et al.,

2010).

Another reason why users performed better with-

out stereoscopy is probably because they had enough

information to determine the position and orientation

of the virtual objects with just one image. The use of

occlusion between virtual objects and the use of shad-

ows as depth cues reduces the impact of conﬂicting

depth cues. In unprepared AR environments, Sands

et al. also showed that stereoscopy did not improve

user performance in target selection tasks when there

is enough information to ﬁnd the position of a 3D

cursor using a 2D visualization method (Sands et al.,

2004).

Despite the fact that there are no statistically sig-

niﬁcant relationship between both factors, we per-

formed a second statistical analysis. We compared

user performance according to the type of visualiza-

tion used ﬁrst. We classiﬁed users into two groups

(stereoscopic visualization ﬁrst, monoscopic visual-

ization ﬁrst). If we take into account the average

performance values, we ﬁnd that in general both

groups of users got better results in their second

round. This improvement is minor for those sub-

jects that started playing with monoscopic graphics,

but is much greater for those who started playing

with stereoscopic visualization. In general, all the

measurements for the ﬁrst and second rounds play-

ing stereoscopic visualization ﬁrst where worse than

the ﬁrst and second round playing monoscopic visu-

alization ﬁrst. Finally users playing with stereoscopy

enabled ﬁrst had worse results with monoscopic vi-

sualization condition in their second round than users

playing the game with monoscopic condition as the

initial round. Playing with stereoscopy enabled in

the ﬁrst round seems to penalize user performance in

the second round, causing worse performance com-

pared to users playing the game ﬁrst using mono-

scopic graphics with no previous experience. This

effect is more noticeable as the task complexity in-

crease, from level one to level three.

Summarizing, our system in monoscopic mode

provides in general enough cues for the user to cor-

rectly relate real and virtual objects. Consequently,

users who started in this mode performed better on

average. Also, the second time they play, using stere-

oscopy, they better interpreted the additional depth

cues. Alternatively, when the user played with stere-

oscopy enabled ﬁrst, she had to deal with the added

complexity of the stereoscopy and the conﬂicting

depth cues. Her performance was not as good until

she played for the second time, in monoscopic mode,

where there are less conﬂicting depth cues and it is

easier to establish spatial relationships between real

and virtual objects. These conclusions were pointed

out by different users after completing their experi-

ments.

We think that our results are application dependent

and the lack of statistically signiﬁcant difference that

let us prove our hypothesis is caused for the small set

and diversity of users that participated in our experi-

ment. Therefore we plan to extend the experiment to

include more users.

7 CONCLUSIONS AND FUTURE

WORK

We have compared user performance when using

stereoscopy in a Desktop AR system with conﬂicting

GRAPP 2012 - International Conference on Computer Graphics Theory and Applications

464

depth cues (viewpoint offset and incorrect occlusions

between real and virtual objects). We studied 32 vol-

unteers that played three different levels of an AR ver-

sion of the Wire Loop game. Each time we recorded

ﬁve performance measurements.

We did not ﬁnd statistically signiﬁcant differ-

ences on user performance under both tested condi-

tions. However, our results show that stereoscopy

does not improve user performance in our Desktop

AR application. Also, users playing ﬁrst with stereo-

scopic graphics found the game easier after switching

to monoscopic graphics. Users starting with mono-

scopic graphics had better results in all performance

measurements.

Users completed a questionnaire after playing the

AR game. They found the game easier to play with-

out stereoscopy, although they said it helped them per-

ceive the virtual objects better.

We plan to extend this study to include more users

to get more data that support the ﬁndings we have de-

scribed here. We also want to evaluate user perfor-

mance when using other AR applications. Finally, we

plan to study how different feedback channels may

provide task related information to a user of an AR

system.

ACKNOWLEDGEMENTS

This research was partially supported by the National

Council of Science and Technology of M

exico as part

of the Special Program of Science and Technology. It

was also funded by grant ALFI-3D, TIN2009-14103-

C03-03 of the Spanish Ministry of Science and Inno-

vation.

Finally, we want to thank Carlos De Jes

us De Barros

for his valuable criticism while preparing this paper.

REFERENCES

Azuma, R., Baillot, Y., Behringer, R., Feiner, S., Julier, S.,

and MacIntyre, B. (2001). Recent Advances in Aug-

mented Reality. IEEE Computer Graphics and Appli-

cations, 21:34–47.

Cakmakci, O., Ha, Y., and Rolland, J. P. (2004). A com-

pact optical see-through head-worn display with oc-

clusion support. In Third IEEE and ACM Interna-

tional Symposium on Mixed and Augmented Reality,

ISMAR 2004, pages 16–25.

Drascic, D. and Milgram, P. (1996). Perceptual issues in

augmented reality. In SPIE Volume 2653: Stereo-

scopic Displays and Virtual Reality Systems III, pages

123–134.

Hou, M. (2003). A model of real-virtual object interactions

in stereoscopic augmented reality environments. In

Seventh International Conference on Information Vi-

sualization, IV 2003, pages 512–517.

Kruijff, E., Swan, J. E., and Feiner, S. (2010). Perceptual

issues in augmented reality revisited. In Mixed and

Augmented Reality (ISMAR), 2010 9th IEEE Interna-

tional Symposium on, pages 3–12.

Lawson, S. W., Pretlove, J. R. G., Wheeler, A. C., and

Parker, G. A. (2002). Augmented reality as a tool to

aid the telerobotic exploration and characterization of

remote environments. Presence: Teleoperators and

Virtual Environments, 11:352–367.

Liu, S., Cheng, D., and Hua, H. (2008). An optical see-

through head mounted display with addressable focal

planes. In 7th IEEE/ACM International Symposium on

Mixed and Augmented Reality, ISMAR 2008, pages

33–42.

Livingston, M. A., Barrow, J. H., and Sibley, C. M. (2009).

Quantiﬁcation of contrast sensitivity and color per-

ception using head-worn augmented reality displays.

In Virtual Reality Conference, 2009. VR 2009. IEEE,

pages 115 –122.

Looser, J., Grasset, R., Seichter, H., and Billinghurst, M.

(2006). OSGART - A Pragmatic Approach to MR. In

Industrial Workshop, ISMAR ’06.

Michael Laforest (2008). Wiiuse - The Wiimote C Library.

http://sourceforge.net/projects/wiiuse. Last checked:

October 7th, 2011.

Nakano, G., Kitahara, I., and Ohta, Y. (2008). Generating

perceptually-correct shadows for mixed reality. In 7th

IEEE/ACM International Symposium on Mixed and

Augmented Reality, ISMAR 2008, pages 173 –174.

Ruling, Z., Hanxu, S., Qingxuan, J., and Fusheng, Y.

(2008). Research on fast and accurate occlusion de-

tection technology of augmented reality system. In

Industrial Informatics, 2008. INDIN 2008. 6th IEEE

International Conference on, pages 111–116.

Sands, J., Lawson, S. W., and Benyon, D. (2004). Do we

need stereoscopic displays for 3D augmented reality

target selection tasks? In Eighth International Confer-

ence on Information Visualisation, 2004., pages 633–

638.

Shidoji, K., Funakoshi, M., and Ogawa, M. (2010). Percep-

tion of absolute and relative distances in stereoscopic

image. Displays, 7524:752418–752418–10.

Sony Computer Entertainment Europe (2009). Inviz-

imals. http://www.invizimals.com/invizimals.php.

Last checked: October 7th, 2011.

Sony Pictures Digital Inc. (2009). District 9 Movie.

Multi-National United Training Simulation.

http://www.multinationalunited.com/training/. Last

checked: October 7th, 2011.

SPRX Mobile (2010). Layar. http://www.layar.com/. Last

checked: October 7th, 2011.

Sugano, N., Kato, H., and Tachibana, K. (2003). The ef-

fects of shadow representation of virtual objects in

augmented reality. In The Second IEEE and ACM In-

ternational Symposium on Mixed and Augmented Re-

ality, 2003., pages 76–83.

STEREOSCOPIC VISION IN DESKTOP AUGMENTED REALITY - User Performance in the Presence of Conflicting

Depth Cues

465