Vis Comput
DOI 10.1007/s00371-009-0333-5
ORIGINAL ARTICLE
Mr-SDM: a novel statistical deformable model for object
deformation
Qizhen He ·Horace H. S. Ip ·Jun Feng ·Xianbin Cao
© Springer-Verlag 2009
Abstract In this paper, we propose a novel statistical de-
formable model (SDM) for Material-related object deforma-
tion, which we called Mr-SDM. In Mr-SDM, by integrat-
ing the prior knowledge of the physical material property
into the training of SDM, we are able to achieve both ac-
curacy and computational efficiency in simulating material
deformation for various applications. Our Mr-SDM train-
ing process takes advantage of the accuracy of Finite Ele-
ment Method (FEM) to generate a set of deforming samples
which enables us to estimate the deformation parameter of
an unknown object based on its material knowledge. Our ex-
periments have shown that Mr-SDM is able to give compa-
rable accuracy with respect to FEM while, at the same time,
reducing the computation cost from O(n
2
) for FEM-based
simulation to O(n) using Mr-SDM.
Keywords Object deformation · Statistical deformable
model · Finite element
Q. He (
) · H.H.S. Ip · J. Feng
Image Computing Group,Department of Computer Science,
City University of Hong Kong, Hong Kong, China
e-mail: rainy@mail.ustc.edu.cn
H.H.S. Ip
e-mail: cship@cityu.edu.hk
J. Feng
e-mail: fengjun@nwu.edu.cn
Q. He · X. Cao
Department of Computer Science and Technology,
University of Science & Technology of China, Hefei, China
X. Cao
e-mail: xbcao@ustc.edu.cn
1 Introduction
Computer simulations are commonly deployed in many ap-
plications, from medical simulation of organ deformation
to industrial CAD/CAM design of different types of prod-
ucts and buildings, as well as in entertainment such as spe-
cial effects for movies. Furthermore, for many of the ap-
plications, real-time or interactive performance of simulated
events is needed. Such demands frequently lead to a trade-
off between accuracy and response time of the simulated
output. To achieve fast response time, e.g. for real-time ani-
mation [1], typically low resolution models and deformation
algorithms that yield fast but less accurate results are used.
While for other applications, such as for computer aided de-
sign [2] and surgical planning [3], that focus on physical
accuracy, high resolution graphical models and computa-
tionally intensive (in both space and time) deformation algo-
rithms are always deployed. Unfortunately, the balance be-
tween the accuracy and efficiency cannot always be met. For
example, in computer aided orthognathic surgery when the
surgeons want to know exactly how the patients look after
the surgery, while, at the same time, they also demand quick
deformation so that they can interactively simulate, compare
and choose among different surgical strategies. In this paper,
we provide a solution to this problem by developing a new
type of statistical deformation model called Material-related
statistical deformable model (Mr-SDM) that integrates the
prior knowledge of the physical material property into the
training of a SDM and use this information for subsequent
deformation simulations. We will first review related work
in Sect. 1.1, followed by presenting the formulation of our
Mr-SDM for model training and deformation simulation in
Sect. 2. The accuracy of the approach for different type of
materials is compared with FEM in Sect. 3 and the paper
concludes in Sect. 4.