Content uploaded by Susanne CJ Boll
All content in this area was uploaded by Susanne CJ Boll
Content may be subject to copyright.
Visualization of Off-Screen Objects
in Mobile Augmented Reality
University of Oldenburg
University of Oldenburg
University of Oldenburg
An emerging technology for tourism information systems
is mobile Augmented Reality using the position and orien-
tation sensors of recent smartphones. State-of-the-art mo-
bile Augmented Reality application accompanies the Aug-
mented Reality visualization with a small mini-map to pro-
vide an overview of nearby points of interest (POIs). In
this paper we develop an alternative visualization for nearby
POIs based on oﬀ-screen visualization techniques for digital
maps. The oﬀ-screen visualization uses arrows directly em-
bedded into the Augmented Reality scene which point at
the POIs. In the conducted study 26 participants explored
nearby POIs and had to interpret their position. We show
that participants are faster and can interpret the position of
POIs more precisely with the developed visualization tech-
Categories and Subject Descriptors
H.5.1 [Multimedia Information Systems]: Artiﬁcial, aug-
mented, and virtual realities
Design, Human Factors, Experimentation
Augmented Reality, Mobile Phone, Orientation
Despite the worldwide ﬁnancial crisis tourism remains one
of the largest economic sectors with about 30% of the world-
wide export business . The results of Morrison et al. 
suggests that digital maps or digital extensions for paper
maps are not the perfect companion for tourists and even
today the two most important tools for tourists are paper
maps and printed tourist guides. Davies et al.  show that
a system’s technical maturity and perfectness is not the most
important aspect. In fact, it’s the interaction a system pro-
vides. Davies’ results suggest that tourists either want a list
of all sights in their environment (”What’s near?”) or de-
tailed information about a speciﬁc object (”What’s that?”).
A promising presentation technique to answer these ques-
tions is mobile Augmented Reality to display information
Copyright is held by the author/owner(s).
MobileHCI’10, September 7–10, 2010, Lisbon, Portugal.
about points of interests (POIs). Due to recent technical
advances mobile Augmented Reality applications for Smart-
phones using a digital compass and GPS became available
to end-users. While the phone’s display shows the camera’s
video a 3D overlay highlights sights in the physical scene.
Applications, such as Wikitude, Layar, or Google’s Goggles
available for the Android platform have each been installed
some hundred thousand times - in a few months.
The augmented scene only provides information about
POIs inside the viewport of the camera. It does not provide
an overview about so-called oﬀ-screen objects - objects that
are outside the viewport because they are besides or behind
the user. Current mobile Augmented Reality applications
(e.g. Wikitude or Layar) for tourists provide the user with
an overview about nearby sights using an additional mini-
map, usually centered in the lower half of the display. The
augmented environment and the mini-map, however, have
diﬀerent reference systems. Therefore, interpreting the 2D
mini-map and align it with the augmented environment de-
mands mental eﬀort.
We assume that an oﬀ-screen visualization directly em-
bedded into the augmentation reduces this mental eﬀort. In
this paper we design a 3D visualization of oﬀ-screen objects
for mobile augmented reality applications. Our aim is to de-
termine if mobile augmented reality applications for tourists
can be improved by replacing the small mini-map with the
developed visualization of nearby objects.
In the remainder of this paper, we present the related work
in Section 2. The used visualization techniques are described
in Section 3. We present the design of the conducted user
study in Section 4. The results are outlined in Section 5
followed by a discussion in Section 6. We close the paper
with a conclusion and outlook to future work in Section 7.
2. RELATED WORK
Providing information about physical objects using mo-
bile devices has received a great share of attention in recent
years. Davies et al. , for example, studied the diﬀerence
between two interaction techniques to acquire information
about nearby POIs. They compared an interaction that en-
ables users to take a photo of a sight and receive according
information (the so-called point-and-shoot interaction) with
an interactive list of POIs in the surrounding. Davies et al.
found that users are happy to use image recognition even
when this is a more complex, lengthy and error-prone pro-
cess than traditional solutions.
Augmented reality for handheld devices is a similar in-
teraction technique that enables users to aim at a physical
object to acquire information about it. Contrary to point-
and-shoot, augmented reality provides instant and continu-
ous feedback to the user. Most research for handheld aug-
mented reality focused on adapting algorithms for mobile
devices which is still an open ﬁeld (see e.g. ). How-
ever, recently Smartphone manufacturers began to include
a compass into their phones. Accompanied by a GPS re-
ceiver augmented reality based on pure sensor data became
Besides providing information by augmenting sights ap-
plications for tourists must also provide an overview about
the environment. Existing mobile augmented reality appli-
cations that are based on GPS and compass data present
a small mini-map side-by-side with the augmentation. The
mini-map is used to present so-called oﬀ-screen objects (i.e.
objects that are not visible in the current camera image
presented on the screen). Visualizing oﬀ-screen objects has
received some attention for other use-cases. Zellweger et
al.  introduced City Lights, a principle for visualizing
oﬀ-screen objects for hypertext. An extension of the City
Lights concept for digital maps is Halo . For Halo circles
are drawn around the object in the virtual oﬀ-screen. Users
can interpret the position of the POI by extrapolating the
circular arc. Baudisch et al. showed that users achieve bet-
ter results when using Halo instead of arrows with a labeled
distance . Burigat et al.  reviewed these results by
comparing Halo with diﬀerent arrow types e.g. by visualiz-
ing distance through scaling the arrows. They found that
arrow-based visualizations outperform Halo, in particular,
for complex tasks. Other oﬀ-screen visualization have been
developed (e.g. Wedge ) but it has not been shown that
these outperform existing approaches.
Oﬀ-screen visualization techniques have also been applied
to virtual environments and augmented reality. Chittaro
and Burigat  compared 2D and 3D arrows as well as map
like techniques to present a single destination in virtual envi-
ronments. They showed that inexperienced users were bet-
ter when using 3D arrows. Tonnis et al.  developed and
evaluated 3D arrows and a map like visualization technique
in an augmented reality application for cars that presents a
source of danger. They conclude from their study that there
is a signiﬁcant advantage for 3D arrows in case of reaction
times over the map like method.
3. VISUALIZATION DESIGN
We assume that a 2D map presented besides the aug-
mented scene, as in Figure 1.b, demands eﬀort to be inter-
preted. We started with the concept of oﬀ-screen objects
and transferred it in 3D to embed the visualization into the
scene. Arrows that directly point at nearby POIs are ar-
ranged around a circle. The centroid of each arrow is located
on this circle and thus, all arrows are on the same plane. The
center of the circle is moved in front of the user’s position to
be inside the viewport. To reduce occlusion among the ar-
rows the plane is slightly tilted towards the user. The arrows
rotate according to the orientation of the phone, just like a
compass with multiple needles. To present the distance be-
tween the viewer and the object the arrows are scaled in
length according to this distance. The scale factor can be
adjusted just like the zoom level in digital maps. As shown
in Figure 1.c, arrows are not hidden if an oﬀ-screen object
becomes an on-screen object (i.e. the object is visible inside
the camera’s video) to avoid confusing the user.
Figure 1: Conception of the visualizations: a) Scene
from top b) mini-map and c) 3D arrows
To compare the arrow based visualization with the state-
of-the-art we implemented a mini-map that also rotates with
the orientation of the phone. A highlighted cone shows the
area of the real world that is visible on the phone’s display.
To obtain comparable results the mini-map is centered at the
same location with the same size as the arrows. To highlight
POIs inside the camera’s video we use circles that overlay the
objects inside the physical scene. If a POI is near the centre
of the display a short description is painted on top of a semi-
ellipse connected to the circle by a thin line. The system
was implemented for Android Smartphones. A screenshot
containing a side-by-side comparison of the application’s two
presentation techniques is shown in Figure 2.
4. USER STUDY
To compare the arrow based visualization of nearby POIs
with mini-maps we conducted a user study with the system
described above. In the experiment participants performed
two tasks using the system. Our assumption was that par-
ticipants are faster with the arrow based technique because
they do not have to mentally align two diﬀerent reference
systems. Thus, we also expected that participants perceive
it as more intuitive. However, we assumed that participants
can localize POIs more precisely with the mini-map because
Figure 2: Screenshot with both visualizations for
illustration purpose. Only one was used at a time
for the evaluation.
of their experience with map usage and because of the more
abstract 2D visualization.
The experiment’s independent variable was the visualiza-
tion technique used to present POIs. In the control condi-
tion, participants used a mini-map and in the experimental
condition they used the arrow based visualization instead.
The study consisted of two tasks. We used a repeated-
measures design for both tasks. The tasked are always per-
formed in the same order but the order of the conditions
have been counterbalanced to reduce sequence eﬀects.
For the ﬁrst task the device displayed four virtual POIs
randomly distributed around the user. Participants’ task
was to read the names of the POIs. In order to do that,
they had to search for the POIs by turning around on the
spot. A POI’s name was written on the top of the screen
if the POI was located at the centre of the display. The
dependent variables were the time needed to read the names
of all POIs and a rating of the visualization technique on a
six point scale.
In the second task the device showed a set of four POIs
all in sight and randomly selected from 12 nearby POIs (e.g.
buildings, shops, and a bus station). Participants were asked
to turn in a speciﬁc direction before starting the task and
memorize the location of the POIs without turning around.
After they ﬁnished memorizing, the device was removed, and
participants had to tell which POIs were displayed. The
dependent variables were the time needed to memorize the
POIs and a rating of the visualization technique on a six
point scale. In addition, the number of correctly estimated
POIs and the diﬀerence between the named POIs and the
displayed POIs in meters and angle were measured.
We set up the evaluation booth on a public place in the
city center of a medium size European city shown in Fig-
ure 3. The study was conducted on a Saturday from 11.00
to 16.00. Two teams of experimenters guided participants
through the tasks simultaneously. Passersby were randomly
asked to participate in the study. After a person had agreed
to participate in the evaluation, the experimenter made the
participants familiar with the concept of presenting POIs us-
ing Augmented Reality and the two visualization techniques.
After conducting both tasks participants were interviewed to
collect demographic information. In addition we asked par-
ticipants to self-assess their experience with Smartphones,
navigation skills, and experience with virtual reality (VE)
on a six point scale.
We conducted the user study with 26 participants, 13 fe-
male and 13 male, aged 21-41 (M=22.4, SD=7.2). The sub-
jects were passersby, so most of them were familiar with the
local place. All sub jects were volunteer, chosen without any
selection by age, nationality or other criteria. None of them
were familiar with the used application.
After conducting the experiment we collected and ana-
lyzed the data. For the ﬁrst task we could not identify sig-
Figure 3: Evaluation at a public place
niﬁcant results. Therefore only the results of the second
task will be discussed in the following. In addition to the
diﬀerences between the visualization techniques, signiﬁcant
eﬀects of gender and stated experience with virtual envi-
ronments are also reported if applicable. Unless otherwise
noted, a t-test is used to derive the p-values.
Using the arrow visualization subjects correctly identi-
ﬁed signiﬁcantly (p=0.023) more POIs (M=2.2) than using
the mini-map (M=1.6). In particular, males were signiﬁ-
cant better (p=0.045) when using the arrow based method
(M=2.1) in contrast to the mini-map method (M=1.3). Fur-
thermore participants who stated to be good in VE were also
signiﬁcant better (p=0,013) when using the arrow method
(M=2.4 compared to M=1.3).
To compare the positions of the displayed POIs with
the positions stated by the participants the respective geo-
coordinates were used. Using these geo-coordinates the de-
viation in meters between these positions were calculated.
Using arrows the distance between the POIs’ correct po-
sition and the guessed position was lower (M=18.0m) than
using the mini-map (M=23.3m) but the diﬀerence is not sig-
niﬁcant. The diﬀerence of the distance from the user to the
correct POI and the distance from the user to the guessed
POI was smaller using arrows (M=29.9m) than using the
mini-map (M=38.8m). However, the diﬀerence was also not
We calculated the angle between the displayed POIs and
the guessed POI. The angular deviation was signiﬁcantly
smaller (p=0.027) using arrows (M=12.4◦) than using the
mini-map (M=20.2◦) In particular, females proﬁted from
the arrows (M=9.7◦) and were signiﬁcant better with ar-
rows (p=0.030) than using the mini-map (M=18.2◦). Also
subjects who stated to be good in VE were signiﬁcant better
(p<0.001) when using the arrows (M=8.5◦) instead of the
On average participants are slightly faster using the ar-
rows (M=21.4s versus M=24.1s) but no signiﬁcant diﬀerence
was found. The ratings are nearly equal with M=2.9 for the
arrows and M=2.9 whereas 1.0 is best and 6.0 is worst.
We found some general tendencies that we, however, can-
not prove to be signiﬁcant. Using the arrows subjects that
stated to have experience with virtual environments were on
average 3.4s slower but identiﬁed 0.3 more objects correctly
than subjects who stated to have little experience. In con-
trast, we found opposite results for the mini-map: Subjects
that stated to have experience with virtual environments
were on average 6.1s faster and identiﬁed 0.5 less objects
correctly than subjects who stated to have little experience.
Furthermore females were always better than males on av-
erage for both visualizations (e.g. 4.9◦smaller angular de-
viation and 0.4 more objects correctly identiﬁed) but slower
(3.3s) in all measured values.
We analyzed our data to ﬁnd anomalies in our sample
of the population. Overall females were 3.8 years younger
than males (p=0.017). Furthermore females rated their nav-
igation skills 0.7 points worse than their male counterparts
(p<0.0001) and females rated their experience with vir-
tual environments 1.3 points worse than males (p<0.0001).
Younger participants rated their own competence worse in
the categories familiarity with the environment (p<0.0001,
r=-0.391), navigation skills (p<0.0001, r=-0.417), and expe-
rience with virtual environments (p<0.021, r=-0.261). An
analysis of covariance (ANCOVA) showed that the younger
females are not the cause why younger participants rated
themselves worse in general because gender is not a covari-
ate in this correlation.
No signiﬁcant diﬀerence between the visualization tech-
niques has been found for the ﬁrst task. We assume that
the data is aﬀected by noise induced by participants lack
of training and the inaccuracy of the used phone’s build-
in compass. However, for the second task the arrows clearly
outperform the mini-map. Participants were faster and more
precise with the 3D arrows. In particular, the absolute
amount of correctly identiﬁed objects was higher.
The results support our ﬁrst assumption, that partici-
pants are faster with the arrow based technique, because
they do not have to align diﬀerent reference systems. Due
to the same reason we expected that users perceive the ar-
rows as more intuitive, but surprisingly all ratings were al-
most equal. One reason might be that subjects were naive
users, which were more interested in mobile augmented real-
ity applications in general than the compared visualization
techniques. Our last assumption, that participants are more
precisely when localizing POIs with the mini-map, was con-
tradicted. We assume that these results are mainly caused
by the improved visualization of directions the 3D arrows
provide. Because of our experimental design we cannot es-
timate if the 3D arrows visualize distances more eﬀectively.
We asked participants to self asses their competence with
virtual environments because we expected that this compe-
tence has a direct eﬀect on participants’ skills to navigate
and orientate in an augmented reality. On average, females
performed always better but with a lower self-assessment
of their experience with virtual environments. This implies
that some males overrated their own competence.
We were surprised that, in particular, subjects who stated
to have experience with virtual environments were sup-
ported by the 3D arrows. This is contradictory to the results
from Burigat and Chittaro , who showed that presenting a
destination in virtual environments using a 3D arrow is espe-
cially suitable for inexperienced users. We identiﬁed two po-
tential reasons causing this contradiction: Let participants
rate their competence in virtual realities might be too impre-
cise, especially compared to the preselected experts Burigat
and Chittaro used in their experiment. Furthermore it’s un-
clear if the results of an experiment about stationary virtual
environments are applicable to mobile augmented reality.
In this paper we described an oﬀ-screen visualization tech-
nique for mobile augmented reality applications using em-
bedded arrows pointing at nearby sights. A system was de-
veloped for Android Smartphones to compare the developed
oﬀ-screen visualization with a state-of-the-art mini-map to
conduct an experiment with 26 passersby in a city centre.
The study showed that 3D arrows enable users to estimate
the position of objects more precisly than a mini-map.
We conclude that existing mobile augmented reality appli-
cations could be improved by using the 3D arrows described
in this work. In addition, we suppose that our results can
be applied to virtual environments in general, so users e.g.
in games might beneﬁt from multi-targeting arrows.
It remains to be examined how the users’ performance
changes with an increasing number of displayed objects
which cause self-occluding arrows. Furthermore an inves-
tigation on possible other variations of oﬀ-screen visual-
izations should be conducted, in particular, the tested ap-
proaches could be uniﬁed in a 3D mini-map.
This paper is supported by the European Community
within the InterMedia project (project No. 038419).
 World tourism barometer excerpt. volume 7, pages
1–7. UNWTO, 2009.
 P. Baudisch and R. Rosenholtz. Halo: a technique for
visualizing oﬀ-screen objects. In Proc. of CHI, 2003.
 S. Burigat and L. Chittaro. Navigation in 3d virtual
environments: Eﬀects of user experience and
location-pointing navigation aids. In International
Journal of Human-Computer Studies, volume 65, 2007.
 S. Burigat, L. Chittaro, and S. Gabrielli. Visualizing
locations of oﬀ-screen objects on mobile devices: a
comparative evaluation of three approaches. In Proc.
of MobileHCI, 2006.
 N. Davies, K. Cheverst, A. Dix, and A. Hesse.
Understanding the role of image recognition in mobile
tour guides. In Proc. of MobileHCI, 2005.
 S. Gustafson, P. Baudisch, C. Gutwin, and P. Irani.
Wedge: clutter-free visualization of oﬀ-screen
locations. In Proc. of CHI, 2008.
 S. Karpischek, C. Marforio, M. Godenzi, S. Heuel, and
F. Michahelles. Mobile augmented reality to identify
mountains. In Adjunct Proc. of AmI, 2009.
 A. Morrison, A. Oulasvirta, P. Peltonen, S. Lemmela,
G. Jacucci, G. Reitmayr, J. N¨
anen, and A. Juustila.
Like bees around the hive: a comparative study of a
mobile augmented reality map. In Proc. of CHI, 2009.
 M. Tonnis and G. Klinker. Eﬀective control of a car
driver’s attention for visual and acoustic guidance
towards the direction of imminent dangers. In Proc. of
 D. Wagner, S. D., and B. H. Multiple Target
Detection and Tracking with Guaranteed Framerates
on Mobile Phones. In Proc. of ISMAR, 2009.
 P. T. Zellweger, J. D. Mackinlay, L. Good, M. Steﬁk,
and P. Baudisch. City lights: contextual views in
minimal space. In Proc. of CHI, 2003.