assembled by Marc Levoy in 1998-99
edited for correctness on January 15, 2000
Q. How did you come up with the idea of digitizing statues?
A.
The idea is not mine. Although my graduate students and I have been developing
3D scanning techniques for several years, we are not the only research group
working in this area. Between our work and that of these other groups,
practical techniques for 3D scanning have been "around" for several years. In
fact, several museums have successfully digitized statues. However, nobody has
digitized a large statue with enough precision to serve as a primary resource
for scientific work, and nobody has tried to digitize a large and coherent
collection of statuary. As my sabbatical year approached, I realized that the
time seemed ripe to mount such a effort. The technology seemed sufficiently
advanced that a successful demonstration might spark some changes in the
methods used by museums to archive works of art and in the methods used by art
historians to study these works.
Q. Why did you select Michelangelo's statues for the project?
A.
Once I latched onto the idea of digitizing statues during my sabbatical year,
and knowing that I wanted my effort to be noticed by museums and art historians
alike, I naturally looked for a body of statuary that is well known and
frequently studied. I also looked for statues that would be feasible to scan.
Michelangelo's statues are not too big, except for the David, not too shiny,
except for the Pieta, and not too geographically dispersed, except for his
statues in Milan, Belgium, and France. I have also been a fan of Michelangelo
since my undergraduate days studying architecture at Cornell University. When
I found out that Stanford University had an Overseas Studies Center in
Florence, the city with the largest concentration of Michelangelo's statues,
the choice became obvious.
Q. Are you scanning all of Michelangelo's statues?
A.
No. We are planning to scan as many of Michelangelo's statues as our time and
resources permit. Some of his statues are not even in Italy. At the least, we
plan to scan the David and the Unfinished Slaves (in the Accademia in Florence)
and the 5 principal statues of the Medici Chapel (in San Lorenzo, in Florence).
We were originally planning to scan the Moses (in St. Pietro in Vincoli, in
Rome), but like many monuments in the eternal city, it is currently under
restoration in preparation for the Year 2000 Jubilee, so it looks as if we will
be unable to scan it this year. Due to the upcoming Jubilee, we will also not
be able to scan the Pieta (in St. Peter's Basilica). We do, however, plan to
scan the architecture of the Medici Chapel, which was designed by Michelangelo.
Finally, we plan to the 1,163 fragments of the Forma Urbis Romae, described in
more detail below.
Q. How does laser scanning work?
A.
A sheet of red laser light illuminates the object, painting a thin stripe onto
it. A video camera looks at the object from the side. By analyzing the
changing shape of the stripe as it sweeps across the object, we can compute the
shape of the object, at least as seen from one side. The resulting data takes
the form of a grid of points in three-dimensional space, sometimes called a
range image. By connecting neighboring points, we produce a mesh of tiny
triangles. By scanning the object repeatedly from different angles, and then
combining the resulting range images, we produce a new, larger triangle mesh
that completely describes the object, at least to the extent we can see it from
the outside. Our triangles are about 1/4 millimeter on a side, small enough to
capture the chisel marks of Michelangelo. After we've finished recording the
shape of the object, we scan it a second time, this time illuminated by white
light, and we use a second camera to record the object's color. To see what
these range images and meshes look like, you can surf to our scan (and 3D fax)
of a Happy Buddha statuette at
http://graphics.stanford.edu/projects/faxing/happy/
(or
http://graphics.stanford.firenze.it/projects/faxing/happy/
in Italy).
Q. How does laser scanning differ from 3D medical scanning?
A.
There are many kinds of medical scanners. Computed tomography (CT) and
magnetic resonance (MR) scanning are two well-known examples. These two
technologies are similar in some ways to laser scanning. All three take
multiple views of the object, and all three create a 3D computer model.
However, lasers see only the outside surface of an object, and must work around
obstructions (what we call occlusions), while CT and MR can see through these
obstructions, allowing a reconstruction of the outside and the inside of the
object. In principle, Michelangelo's statues could be scanned with CT or MR.
However, CT and MR scanners and big, expensive, and not very portable.
Ironically, as big as they are, they are not big enough to scan any of
Michelangelo's statues. Even if you removed his statues from their bases, you
couldn't fit them through the circular hole in the center of the scanner. I
might add that our spatial resolution is higher than most CT or MR machines,
and our scanning techniques also record color, which CT and MR ignore.
Q. What's new in this project?
A.
Laser scanning has been around for a while. Techniques for combining multiple
range images are relatively new. The particular techniques we are using in
this project were invented by us a few years ago for our 3D faxing project (see
above). What else is new? Nobody has ever scanned such large statues before.
This forced us to design our own scanner and motorized gantry, which you can
see on our web page. Although people have scanned larger objects, like ships,
and people have scanned objects with resolution as fine as ours (0.25mm),
nobody has ever scanned such large objects (The David is 23 feet tall on his
pedestal) at such fine resolution. This combination of size and resolution
will make our datasets two orders of magnitude (that's 100x) larger than almost
any 3D computer model in existence. Finally, nobody has ever scanned the
complete contents of a museum before. We are scanning the entire central suite
of the Accademia Gallery - architecture, wall decorations, paintings, and
statues - and the entire New Sacristy in the Medici Chapel, right up to the top
of its dome.
Q. How big will your datasets be?
The raw data we are acquiring for each statue ranges in size from 16 gigabytes
for a small statue to 30 gigabytes for the David. Our final computer models
will range from about 8 to 15 gigabytes. Measured in triangles, our raw
datasets range from 500 million triangles to 2 billion triangles, and our final
triangle meshes will range from 250 million to 1 billion triangles. We can
simplify these models, of course, for example by removing triangles in
relatively flat areas. Using this technique, we can make the models as small
as you like, but they'll look correspondingly coarser. We can even compute a
reasonable-looking model of the David that you could store on today's personal
computers. However, it wouldn't rotate very fast on the screen.
Q. Does the quality of the available light affect the precision of the laser scanning?
A.
It does not affect the laser scanning, but it does affect the color scanning.
The museums will be reducing their ambient lighting level during our project.
Interestingly, turning off the electric lights in the Medici Chapel shows us
(and a few lucky visitors) the space as Michelangelo intended us to see it,
with light entering only from a few windows high in the walls and at the top of
the the dome. This light, bouncing off the walls of gray Pietra Serena
limestone as it filters down into the chapel, fills the space with a sombre
blueish light that is in keeping with its funerary purpose. Amidst the general
gloom, the highly polished figure of Night appears supernaturally white, as if
bathed in moonlight - another part of Michelangelo's design. It is a moving
experience to stand alone in the Chapel when it is lit this way.
Q. How is it that art historians could consistently have mis-measured the height of the David?
A.
I don't know how it happened, but I can hazard a guess. It's hard to measure
the David; he is very tall. I suspect that each art historian trusted the
height given by an earlier art historian. Somebody long ago made a mistake,
and the mistake proliferated. Even the classic 5-volume study of Michelangelo
made in the 1950's by Charles De Tolnay is wrong. The only book I have ever
found that gives the correct height is the catalogue of the Gipsoteca, a
little-known collection of casts of famous Italian statues. This collection,
now stored in a warehouse on the outskirts of Florence, was made in the
mid-1800's when Florence was briefly the capital of a newly united Italy.
By the way, that collection includes a full-size plaster cast of the David.
Q. Why do you want to capture Michelangelo's chisel marks?
A.
On his finished or nearly finished statues, Michelangelo often left the surface
deliberately bumpy. The tiny shadows cast by these bumps deepen the shading of
curved surfaces, like muscles. If we want our computer models to look
realistic, we have to capture these bumps. On his unfinished statues, like the
"Slaves", our ability to capture chisel marks is even more important. Part of
their appeal lies precisely in the fact that they are unfinished. In these
statues we can see the master at work: we can see which tools he used, we can
admire the extraordinarily long and deep furrows he made with his chisels -
something that astonished his contemporaries, and we can see the order in which
he penetrated into the block seeking the forms within. On these Slaves at
least, he worked inwards from one or two sides only, not from all around as
most sculptors do. One problem that art historians argue about endlessly is
the number of teeth on Michelangelo's chisels. Starting from our computer
models, we think it might be possible to write a computer algorithm that could
analyze the distance between adjacent furrows and then draw a map of the statue
showing for example where he used a two-tooth chisel and where he used a
three-tooth chisel. Trying to answer a question like this without our models
would require extremely precise hand measurements repeated thousands of times
over the entire statue.
Q. How do art historians based at universities see themselves using the final product?
A.
Many art historians have inquired about our data, but like any new technology,
they are not exactly sure how to use it. Among Michelangelo scholars, one
inquiry I have received that particularly struck me was posed by Professor
Irving lavin of the Institute for Advanced Study in Princeton. He wants help
with a long-standing problem that involves the exact dimensions of the
allegories on the Medici tombs. We are currently scanning these statues, we
hope to be able to help him. I should note that in the short term, it will be
difficult for most art historians to use our models simply because they are so
large. However, we will eventually develop simplified models, our software for
viewing and analyzing these models will improve, and computers will get faster.
Q. Will your computer models change the way that people view art?
A.
Perhaps. To give a simple example, museum visitors see most statues from a
limited set of viewpoints. Computer models give people a way to look at
statues from any viewpoint. They can also change the lighting. In the case of
Michelangelo's statues, most of which are large, views are always from the
ground looking up. Michelangelo knew this, and he designed his statues
accordingly. Nevertheless, it is interesting and instructive to look at his
statues from other viewpoints. Looking at the David from unusual directions
has taught me many things about the statue's ingenious design. I describe a
few of these in the photo essay at
http://graphics.stanford.edu/projects/mich/head-of-david/.
Q. Do David's two eyes really diverge? Is this a flaw in the statue?
A.
They do diverge, but it's almost certainly not a flaw. It is widely
accepted that Michelangelo introduced distortions into his statues in order to
achieve certain visual effects. In the David, his head and right hand are too
large for his body. Also, his famous furrowed brow actually protrudes from his
forehead in a way that is anatomically impossible. In the Pieta in in
St. Peter's Basilica, if the Virgin were to stand up, you would see that she is
mostly legs. Also, her right arm, which supports her son's shoulders, is too
long. These are some of the tricks Michelangelo used in order to gracefully
depict a grown man lying in a woman's lap. Returning to the David, I first
noticed that his eyes diverge while examining the statue from atop a
scaffolding. I shot this
photograph to document my
observation. Later on, by studying these
renderings of our 3D computer model, I came up with a possible explanation
for the divergence; Michelangelo was adjusting the location of the two pupils
in order to optimize the appearance of the two profiles of his head. These
profiles look surprisingly different from one another, as the renderings show.
Of course, I'm a computer scientist, not an art historian, so this hypothesis
is really just an educated guess.
Q. Will computer models allow real museums to be replaced by virtual museums?
A.
Unlikely. No model displayed on a computer screen can replace the experience
of walking around a statue. Head-mounted displays and similar virtual reality
systems have made good progress in the last decade, but they still do not
replace being there. On the other hand, they can enhance being there.
For example, we have been scanning in the museums of Florence while they are
open to the public. One day, while I was examining a scan of Michelangelo's
St. Matthew on my computer screen, a visitor standing behind me turned to her
companion and exclaimed that the chisel marks seem so much clearer and more
dramatic on the screen than on the statue. At the time, I was looking at
St. Matthew's cheeks, in closeup, and using a virtual light that just raked the
surface. Both the viewpoint and the lighting are different from what a visitor
sees. Pretending not to hear her comment, I slowly circled the light around
the mesh. "Wow", she added, "if only we could do that on the real statue!"
Her comment is revealing. On the surface it seems ludicrous to place a
computer in front of a statue, and on the computer screen to display a 3D model
of that same statue. In reality the computer focuses the visitor's attention
on the statue, and it allows them to view it in a new way. If we allowed
visitors to manipulate the models themselves with a joystick, they would surely
jump at the chance. By exploring the statue themselves, or by changing its
virtual lighting, they make the viewing of art an active rather than a passive
experience. The art museum becomes a hands-on museum.
Q. How to you plan to virtually reconstruct the smoothed-out beard of the Moses?
A.
I don't know if this is possible, but we have a few ideas. One approach might
be to measure the depth and shape of the curls at locations that have not been
worn away and then instruct the computer to morph the surfaces of the worn
curls until they take on the same basic profile as those in the pristine areas
of the statue. Another possible approach might be to run the 3D equivalent of
an edge sharpening operator, such as you might find in Photoshop, over our
computer model of his beard. A more sophisticated approach would be to create
an accurate physical model of the process that wore down the features in the
first place, and then reverse it. Unfortunately, it now looks as if we will be
unable to scan the Moses this year because, like many monuments in Rome, it is
currently under restoration in preparation for the Year 2000 Jubilee.
Q. Who owns the final digital product?
A.
This question, and many like it, is not entirely resolved. In fact, the
exploration of intellectual property issues is one of the ways in which this
project promises to break new ground. First, what do you mean by "owns"? It
is misleading to talk about copyright; the more important questions are about
control and about royalties. Second, what do you mean by "digital product"?
We could talk about computer models - the set of polygons, defined by X,Y,Z
coordinates and color, that record the shape and reflectance of the surface of
the statue, or we could talk about computer renderings - images computed from
the models. Since the goals of our project are strictly scientific, our
arrangement with the museums is simple and flexible. We are allowed to use
and, to a limited extent, distribute our models and computer renderings for
scientific use only. In the event we, or they, desire to use the models
commercially, there will be further negogiations and probably the payment of
royalties. Needless to say, the museums and I are working closely together to
find reasonable and fair solutions to these questions. Both of us realize that
because of the high visibility of this project, the solutions we find may set
precedents for others.
Q. So are you planning to make copies of the David?
A.
Virtual copies, yes; physical copies, no. Some of the articles about the
project that have appeared in the lay press misrepresent us completely on this
point. This is a scientific, not a commercial project. We are not planning
to, nor are we permitted to, make physical replicas of the statues of
Michelangelo.
Q. Will this technology advance to the point that the major sculptures throughout the world will be digitized?
A. Not anytime soon. The technology is still difficult to use. The museums of the world are just now beginning large-scale digitization of their graphical (two-dimensional) artifacts. However, if there are compelling economic reasons to scan a particular statue, for example in order to make copies for sale to the public, then it will be done.
Q. How big is the map?
A.
The map is 22 meters x 15 meters. It is divided into rectangular slabs whose
thickness varies from 3cm to almost 10cm. Some fragments are as large as are
only a few centimeters long, others are more than a meter long and weigh
several hundred kilograms.
Q. Where is the map?
A.
In 200 A.D., the map graced the back of the Roman census office, officially
called the Templum Sacrae Urbis. The map has had several homes over the
centuries. At one point, pieces of it were broken up and used as rubble to
build a garden wall in the Farnese Palace in Rome. At present, all known
fragments reside in the City of Rome's civic history museum. However, at this
moment, they are packed in crates awaiting completion of a restoration of the
museum.
Q. How much of the map still exists?
A.
Only 15% of the map's fragments remain. However, experts believe that most of
them are likely fit together due to the unique way the map fell from the wall
on which it was mounted and the way it was buried by other rubble.
Q. How many pieces have already been fitted together?
A. There are 1,163 fragments altogether. Of these, fewer than 200 have been identified (and in some cases fitted together with other pieces). Another roughly 500 have not been identified, and the rest have no incised lines, which makes identification impossible, at least for a person. However, the lack of incised lines does not make fitting impossible. Such fitting could be based on 3D shape, marble veining, and other cues that only a computer can take advantage of.
