Restoring 3D Digital Woodcut Shape for Reproducing Ancient Book

1. Introduction
Cultural heritage is important for studying history, and
some studies have been carried out to preserve it as
digital data. Hanpon are ancient woodcut-printed books,
all of which are regarded as important cultural heritage
in Japan. (see Figure 1 for woodcut and hanpon). The
woodcut looks black as its surface is covered with ink.
A lot of studies about hanpon and woodcut have been
done. However, there is little study to connect these two
studies using digital techniques. The main contribution
of our study is developing digital techniques for studying
hanpon history using Woodcut information.
Fig. 1 Woodcut and hanpon.
There are some versions of hanpon whose true published
time is not clear. As the paper or woodcut have
shrunken over time, it is possible to guess which version is published early and which is late. Recently, some
woodcuts were found, and we hope to know the original
size of hanpon printed by these woodcuts. If we could
do so, we can know which version of hanpon is printed
by the founded woodcuts. Warped in shape, however,
they cannot be utilized to print hanpon directly. To solve
this problem, we propose techniques to restore woodcut
shapes in a digital way so that we can reproduce ancient
hanpon. As traditional method, a printer put some water
on the back of Woodcut and the Woodcut shape can
be restored near a plane. But the size of the Woodcut
restored using the traditional method is different to the
original one. To solve this problem, a digital technique
for restoring the shape of digital Woodcut is proposed.
Then we can produce original ancient hanpon.
Basically, our system is a virtual printing system. Okada1
proposes a system for sculpturing and printing in a virtual
world. Focusing on virtually sculpturing 3D objects,
his system can print Japanese drawings if one paints
some colors on the virtual surface of a carved object.
Yet, the system uses a carved plane for virtual printing
and cannot be utilized in the case of measured woodcut
digital data, as real woodcuts are distorted a little. For
virtual printing on the distorted surface of woodcut,2 uses
a small plane to fit to a distorted woodcut surface and
make a virtual printing. However, this method ignores
changes in woodcut size. To acquire a hanpon the same
as the ancient one, it is necessary to restore woodcut
shapes.
Fig. 2 Aligned 3D digital Woodcut model
For restoring woodcut shapes, we can learn a lot from
some research concerned with wood drying.3
Finite element
method (FEM), for example, is utilized to simulate
moist translation and shape variation. This study in the
wood research field mainly focuses on how to decrease
shape distortion that occurs during the drying process.
As wood is an orthotropic material, we propose an algorithm
for determining orthotropic direction to restore
woodcut shapes.
The following is the procedure to restore the woodcut
shape for reproducing the ancient book of hanpon. First,
using commercial software developed from the algorithm4
, we measure 3D of the woodcut, whose data are
to be aligned to make the 3D digital woodcut. After the
woodcut shape is restored, we can print out virtual hanpon.

2. Measuring and aligning 3D Woodcut
point cloud data
The non-contact 3D-Measurement machine VIVID910
is utilized to measure woodcut point clouds. When measuring
the woodcut, it is necessary to take bump patterns
on the woodcut surface as precisely as possible. Because
the measurement machine cannot measure all the surface
of the object at once, we need to generate object shape
models by using measured point clouds from multiple
viewpoints. Commercial software is utilized to align
cloud data. At first, the rough corresponding point is
defined on different cloud data. Then, the software can
value the globe errors of the total aligned point clouds
and adjust the position and direction of these point
clouds. At the end, the point clouds are aligned in one
3D digital woodcut model. Figure 2 show the aligned result.
Different colors show different parts of the woodcut
cloud point data.
3. Constructing wood board model
As the woodcut surface is very delicate, the computing
cost will be considerable if one uses this delicate model.
A rough mesh model is constructed to restore the woodcut
shape. As the woodcut is made from a piece of wood
board, if the wood board shape is restored, so will be
the woodcut shape. Hence, the rough model is the wood
board. This wood board model will is a box after its
shape is restored.
To construct the wood board model, we utilize the woodcut
section. In the Figure 3, the bottom line is the woodcut
model section, and the top line is the constructed
wood board model section. The top line is the connection
line between local highest points of woodcut model
sections. Some woodcut sections are utilized to construct
the total wood board model. The distance between these
sections is the same as the woodcut's thickness. After all sections are processed with this method, we can obtain
the wood board for restoring.
Fig. 3 Constructing the wood board model from Woodcut
section
4. Restoring the wood board model
As mentioned above, wood is an orthotropic material.
The wood study needs to consider the three main axes
of the wood. Figure 4 shows a tree trunk form in a coordinate
system with the three standard directions: the
trunk axis or fiber direction L, the radial direction R that
passes through the tree’s core, and the tangent direction
T along the annual ring. When wood dries and shrinks,
shrinkage in the tangential, radial, and fiber directions
occurs in a ratio of 10:5:0.5. With this reason, the wood
board distorts.
While the surface of woodcut is black, on the woodcut
end, we can find some exposed parts, including the annual
ring pattern. From this annual ring, the tree’s core can
be understood as the annual ring is nearly a concentric
cylinder. From the position of the tree’s core, the wood
three standard directions can be determined all over the
wood board.
Fig. 4 Wood three axis.
The woodcut is made of wild cherry tree or boxwood.
Shrinkage of this type of the wood is about 0.31% on
the tangent direction T , and 0.17% on the radial direction
R. Since the shrinkage on fiber direction L is very
small, its change L is ignorable. If the moisture content
is lower than 30%, the wood starts distorting, in Japan,
after the moisture content is 15%, the moisture content
stops changing, and the shape stops distorting. As Figure
5 shows, point A is one point in the wood board. ε is the
vector to show plastic strain on the point A. T is tangent
direction and R is radial direction. The plastic strain can
be shown as the following equation:
ε = εT
− εR =(ST
× T − SR × R) × ΔW (1)
where, ST is shrinkage on direction T and SR is shrinkage
on direction R. ΔW is moisture content variation and is
1%. Using this equation, the plastic strain ε can be computed.
The restoring process is repeated until the wood
board surface gets closer to a plane or the total moisture
content reaches 30%. This is the restored wood board.
As the connection between the woodcut point cloud and
wood board mesh is known, it is easy to obtain a restored
woodcut point cloud from the restored wood board. As
the local highest point is on a plane after shape restored,
the virtual hanpon can be printed easily by using the
height information of the restored woodcut. If the height
is higher than 0.3mm under the highest point, it is the
part for printing. The printed result is shown in Figure 6.
Fig. 5 Plastic Strain.
5. Results
The final rendering result is carried out on a computer
with the GPU (Graphics Processing Unit) and can render
the hanpon on real time. The graph card is NVIDIA
GeForce 6800 GS.
Since the Japanese paper became old and was not as
white as it was printed out, brown color captured from
old Japanese paper is added into the virtual printing result.
The virtual printing result and the Japanese fiber model are utilized to render based on the fiber reflection
model.5 Figure 6 shows the final rendering results. The
left image shows the hanpon appearance with white
color after it was printed out about 200 years ago. The
right one shows the result after hanpon color variation.
Using the technique proposed, the ancient hanpon is represented
in the virtual world.
As the woodcut shape is restored, the size of printed
hanpon is the same as the original hanpon when it was
printed hundreds years ago. Comparing the size of hanpon
and virtual printed result using proposed method,
the published history of hanpon can be understood well.
This information is very important for history research.
Fig. 6 Rendering result of hanpon.
6. Conclusion and discussion
This paper proposes the techniques to restore woodcut
shapes and reproduce ancient hanpon. The bump pattern
of woodcut surface is very minute, and the color of
woodcut is black. To restore the woodcut shape, therefore,
the FEM method is utilized. This method can obtain
fine printing results even woodcut is distorted. To
our knowledge, it is first time to try to study hanpon history
using virtual Woodcut printing method.
Some work, however, needs to be done in the future.
Right now, using 3D scanner cannot capture the woodcutfs
details. Thus, we need to develop techniques to improve
original 3D data precision. One idea is to use the
high resolution camera to take high resolution photos in
a different lighting environment. Using image processing
technique makes the printed areas possible to be
extracted. Then, using the proposed techniques in this
paper to calibrate the hanpon size, we can obtain delicate
printed results. We also hope to develop a Virtual Reality
system in which the user can watch the cultural heritage
and touch its surface at the same time. This Virtual Reality
system is not only a new type digital museum, but
also is entrainment system such as game as well.
References
1) M.Okada, S.Mizuno and J.Toriwaki: Virtual Sculpting
and Virtual Wood block Printing by Model-Driven
Scheme, the Journal of the Society for Art and Science,
Vol.1, No.2, pp.74–84 (2002).
2) Yin, X., Eto, T., Akama, R., Nagai, K. and Tanaka,
H.T.: Digital Woodcut Measurement and Ancient Hanpon
Rendering, Proceedings of 2008 ASIA GRAPH,
Vol.2, No.1, pp.31–36 (2008).
3) Carlsson, P. and Esping, B.: Optimization of the wood
drying process, Structural and Multidisciplinary Optimization,
Vol.14, No.4, pp.232–241 (1997).
4) Soucy, M. and Laurendeau, D.: A General Surface Approach
to the Integration of a Set of Range Views, IEEE
Trans. Pattern Anal, Vol.17, No.4, pp.344–358 (1995).
5) Yin, X., Cai, K., Takeda, Y., Akama, R. and T.Tanaka,
H.: Measurement of Reflection Properties in Ancient
Japanese Drawing Ukiyo-e, Proceedings of 8th Asian
Conference on Computer Vision (ACCV 2007), Part I,
LNCS 4843, pp.779–788 (2007).