line detection is the perfect candidate to be integrated
in an electronic orientation table.
In figure 8, there are some visual results. We have
zoomed on some examples of mountains to show in
detail the obtained registration. The two original reg-
istrations using only the DEM and the smartphone lo-
calization objectively completely fail but the result is
better by coupling DEM and image processing.
5 CONCLUSION
This paper presented a robust two-steps method able
to create a smartphone application identifying instan-
taneously the skyline in an image. The proposed
method starts by image simplification based on effi-
cient color difference between aligned pixels giving
rise to a score map between sky and non-sky pix-
els. Then, based on available data directly from the
smartphone offering a coarse localization, a second
step matches extracted skylines from the digital ele-
vation model with this map to identify precisely the
real skyline. The interest of this couple between im-
age processing and digital elevation model is twofold:
it gives an efficient tool in this context of electronic
orientation table, and it allows to improve augmented
reality tools based on smartphones using image pro-
cessing techniques. Moreover,the tagged dataset built
in this paper will benefit the entire community in
this field. Future works will first consist in analyz-
ing robustness to meteorologic conditions and then in
detecting other notable elements in the image using
image analysis but also the DEM. Let us remember
that our final goal is to present to the user informa-
tion concerning points of interest in the image using
augmented reality. For this, we also intend to use
databases containing monuments, roads, etc.
There is still an immense potential in this field be-
cause of its utility in many areas like in tourism of
course but also, for example, for hikers who want to
visualize hiking trails, for persons interested in geol-
ogy because it would be possible to show them the
structure of the ground also and so on. Smartphones
are today abundant in our daily life and augmented
reality systems could be developed by mixing their
instruments with image processing algorithms.
ACKNOWLEDGEMENT
The authors acknowledge the support from Le Pro-
gramme Avenir Lyon Saint-Etienne Image et Percep-
tion Embarquées (PALSE IPEm – ANR-11-IDEX-
0007). They also would like to thank Thierry Joliveau
for his help in this work.
REFERENCES
Ahmad, T., Bebis, G., Nicolescu, M., Nefian, A., and Fong,
T. (2015). An edge-less approach to horizon line de-
tection. 2015 IEEE 14th International Conference on
Machine Learning and Applications (ICMLA), pages
1095–1102.
Baboud, L.,
ˇ
Cadík, M., Eisemann, E., and Seidel, H.-P.
(2011). Automatic photo-to-terrain alignment for the
annotation of mountain pictures. In Proceedings of the
2011 IEEE Conference on Computer Vision and Pat-
tern Recognition, Washington, DC, USA. IEEE Com-
puter Society.
Bellet, A., Habrard, A., and Sebban, M. (2013). A survey
on metric learning for feature vectors and structured
data (arxiv:1306.6709v3). In Tech. report.
Byung-Ju Kim, Jong-Jin Shin, H.-J. N. and Kim, J.-S.
(2011). Skyline extraction using a multistage edge fil-
tering. International Journal of Electrical, Computer,
Energetic, Electronic and Communication Engineer-
ing.
Fedorov, R., Fraternali, P., and Tagliasacchi, M. (2014).
Mountain peak identification in visual content based
on coarse digital elevation models. Proceedings of
the 3rd ACM International Workshop on Multimedia
Analysis for Ecological Data, pages 7–11.
Hung, Y.-L., Su, C.-W., Chang, Y.-H., Chang, J.-C., and
Tyan, H.-R. (2013). Skyline localization for mountain
images. In 2013 IEEE International Conference on
Multimedia and Expo (ICME), pages 1–6.
Lie, W.-N., Lin, T. C. I., Lin, T.-C., and Hung, K.-S. (2005).
A robust dynamic programming algorithm to extract
skyline in images for navigation. Pattern Recogn.
Lett., 26(2):221–230.
Peak.AR (2010). http://peakar.salzburgresearch.at/. Ac-
cessed: 2016-09-05.
Peakfinder (2016). https://www.peakfinder.org. Accessed:
2016-09-05.
Peakscanner (2016). http://www.peakscanner.com/. Ac-
cessed: 2016-09-05.
Perrot, M., Habrard, A., Muselet, D., and Sebban, M.
(2014). Modeling perceptual color differences by lo-
cal metric learning. In European Conference on Com-
puter Vision (ECCV), pages 96–111. Springer Interna-
tional Publishing.
pointdevue (2015). https://itunes.apple.com/fr/app/
point-de-vue/id341554913?mt=8. Accessed: 2016-
09-08.
Saurer, O., Baatz, G., Köser, K., Ladický, L., and Pollefeys,
M. (2016). Image based geo-localization in the alps.
Int. J. Comput. Vision, 116(3):213–225.
swissmap (2016). https://itunes.apple.com/fr/app/
swiss-map-mobile/id311447284?mt=8. Accessed:
2016-09-08.