You are using an outdated browser that does not fully support the intranda viewer.
As a result, some pages may not be displayed correctly.

We recommend you use one of the following browsers:

Full text

International cooperation and technology transfer
Mussio, Luigi

> computer images, if a fairly accurate rendering software is
Then, the number of the required panoramic views is set up: it
can range from a few, to get a plan of the environment, until a
dense distribution of photographs, to build a more detailed 3D
virtual model.
A good planning is an important issue for realizing virtual
visits, mainly due to following reasons: 1) improvement of the
on the field data collection step, 2) time reduction to convert the
data in digital format, 3) optimization of camera stations setup,
accor-ding to the working environment.
The shootingt is the most important step of the realization of the
virtual movie. In order to meet the previously exposed mathe
matical constraints, the optical centre of the camera must be
aligned with its rotation point. This can be carried out with a
particular tripod, adjustable at different positions. High
accuracy (in order of millimeter) for the optical centre position
is required only for very close objetcs. In this case slight
misalignement can lead to changes of visible portions of the
same object on conse-cutive frames. Dealing with environments
where objects far and close from camera don’t occur, the focus
setup, so as the check for alignement, become less problematic.
For instance, if we see the same background behind the edge of
a book, positioned in front of the camera, looking and moving
the camera, thus the camera is aligned.
The final result of building virtual visit is affected by camera
objective as well. Choosing a wide-lens (max. focal length of
28mm.) allows esier movements in vertical direction in the final
movie, on the other hand the fish-eye lens allows to map the
infinity (180°) on the image plane, but its strong geometrical
distortions have to be to corrected prior to compose the pano
ramic views.
The camera is mounted vertically to take advantage of the view
camera angle. The number of photographs, needed to cover a
360° angle (see fig. 3), must be taken in such a way that the
overlapping zone between two adjacent photographs is about
30%. This allows to the algorithm, that produces the panoramic
view, to calculate the right correlation between two images and
to determine its best overlapping area.
Figure 3
Starting from the images acquired with a scanner, in order to
build a panoramic view the STITCH algorithm is employed.
Given the relationships (7) and (9) relating consecutive images,
it is possible to perform an image mapping on any surface, on
which a 2D reference system is established. In our case we use
a cylindrical surface: this can be considered as a plane once the
lateral surface of the cylinder has been unrolled.
To build the image texture on this surface each photograph is
projected onto the cylinder, calculating the focal length and the
camera viewangle used for each snapshot. The mapping algori
thm needs an initial estimate of the camera viewangle, then it
provides the optimal angle that gives the highest correlation
between each images pair. Due the time consuming of the algo
rithm, it is advisable to test only two sample images and then,
on the basis of the results, to apply the algorithm to all the
photo-graphs. Among the parameters improved in the test we
remember the following:
> Horizontal offset between images
> Vertical offset between images
> Final height of panoramic view
> The best camera viewangle
The algorithm produces a cylindrical map pasting one after the
other all the photographs. Note in fig. 4 - 7 the different results
between a manual image pasting and that performed by the
In fig. 8 and 9 some results of image processing regarding the
"Ancient Courtyard" of the University of Padova are showed.
This technology was used to build a virtual visit of the "Ancient
Courtyard" and of the Anatomical Theatre of the University of
Padova. To the aim several camera stations were set up,
namely: five inside of the Ancient Courtyard, two on the
staircases that lead to the Courtyard and eight inside of the
Anatomical Theatre.
Some camera stations are resulted open, i.e. only a portion of
entire environment could be viewed fom these positions, there
fore only the more meaningful photos were taken. For some
others stations the corresponding images were not used , due
the overexposure of the photographs 3 .
Anyway the results obtained for the Ancient Courtyard are
excellent; these allow an interactive navigation in the courtyard,
that is displayed to the user just as it is.
In the Anatomical Theatre the image capture was difficult, not
only due to the equipment setup, but also because, being a
narrow and tall place, a lens with a very short focal length
would have been more useful.
The result however is good, although a visitor could desire
more movement capabilities in vertical direction along the 3D
virtual model.
3 As the panoramic view results by the composition of two
separate images, to avoid to let recognize the overlapping area
the same light conditions in adjacent photographs have to be
preserved. For this reason the camera diaphragm can be modi
fied very slight, making useless those photographs taken in
back lighting condition.