ORIGINAL ARTICLE
Y.-Z. Chang (*)
Department of Mechanical Engineering,
Chang Gung University, Tao-Yuan 33302, Taiwan
e-mail: zen@mail.cgu.edu.tw
Z.-R. Tsai
Department of Computer Science and Information Engineering
Asia University, Taichung, Taiwan
S.-T. Lee
Medical Augmented Reality Research and Development Center
Chang Gung Memorial Hospital, Linkou Branch, Tao-Yuan., Taiwan
This work was presented in part at the 13th International Symposium
on Artificial Life and Robotics, Oita, Japan, January 31–February 2,
2008
Artif Life Robotics (2008) 13:242–245 © ISAROB 2008
DOI 10.1007/s10015-008-0532-6
Yau-Zen Chang · Zhi-Ren Tsai · Shih-Tseng Lee
3D registration of human face using evolutionary computation and
Kriging interpolation
1 Introduction
Registration of 3D geometric data coming from multimodal
images is essential for exploiting the complementary infor-
mation between them. In medical imaging, the data can
be obtained from computer tomography (CT), laser range
finder, or magnetic resonance imaging (MRI), and the
information can be used in image-guided procedures, such
as positioning for frameless neurosurgery.
The Iterative Closest Point algorithm (ICP), together
with the K-D search method (K-D tree), has become a
popular registration scheme. Approximate K-D tree search
algorithm (AK-D tree) was proposed by excluding the
backtracking in K-D tree, which improves runtime effi-
ciency with the sacrifice of reducing the correspondence
accuracy. However, these conventional schemes are
extremely sensitive to initial trial pose and requires multi-
ple trials to find a reliable solution. Besides, the scheme
demands huge computing power when large data set is
involved in either reference image or template image, which
is not uncommon in medical applications. We thus pro-
pose a scheme that is both consistent in each run and fast
enough to extend the applicability of 3D registration in
intra-operative applications.
To illustrate the validity and applicability of the pro-
posed approach, a problem composed of 174 635 points
computer tomography (CT) reference image and a 11 280
points template image, derived from a laser range finder, is
provided.
2 The proposed method
The proposed registration scheme is composed of a coarse
transformation stage and a fine-tuning stage.
2.1 The coarse transformation stage
In the first stage, fuzzy c-mean (FCM) is used to reduce
the data amount of template 3D image, and evolutionary
Abstract This paper proposes a fast and robust 3D human
face geometric data registration strategy dedicated for
image-guided medical applications. The registration scheme
is composed of a coarse transformation stage and a fine-
tuning stage. In the first stage, fuzzy c-mean is used to
reduce the data amount of template 3D image, and evolu-
tionary computation is implemented to find optimal initial
pose for the Iterative Closest Point plus k-dimensional (K-
D) tree scheme. In the second stage, the huge reference
image data are replaced by a Kriging model. The time-con-
suming search for corresponding points in evaluating the
degree of misalignment is substituted by projecting the
points in the template image onto the model. To illustrate
the validity and applicability of the proposed approach, a
problem composed of 174 635 points reference image and
an 11 280 points template image is demonstrated. Compu-
tational results show that our approach accelerates the reg-
istration process from 1361.28 seconds to 432.85 seconds
when compared with the conventional ICP plus K-D
tree scheme, while the average misalignment reduces from
11.35 mm to 2.33 mm.
Key words Human face registration · Evolutionary com-
putation · Kriging model · Iterative closest point · Image-
guided therapy
Received and accepted: July 18, 2008