Reconstruction
of Human Face
in 3D
Environment
Pravin
Yadav
Symbiosis International
University, Lavale, Mulshi, Pune 411 042, INDIA.
E-mail: pravinyadav@gmail.com
Key words: 3D face, physically based system, facial Reconstruction, facial
modeling, anthropometry, anatomy, forensic science, tissue-depth thickness, face
approximation
Abstract
In this paper, Researcher presents flexible and dynamic methodology to
reconstruct human face which can be quickly adopted in any 3D environment. Proposed
methodology is unifying approach that deals with translated inputs from various
streams like anatomy, forensic science, and anthropometry. Proposed system imitates
the physically based system that relies on anthropometric dimensions and tissue
depth information. Proposed system also covers morphometric shapes along with
anthropometric dimensions for all facial features. It is phase wise development
approach (Section 2) and reinforced with concepts like Jeevak and Vyaktirekha (Section
2.2) which makes it faster and flexible.
1. Introduction
The whole world population is increasing by
every second. Within approximately 9”x 7”x 8” volume, we can see billions of variations.
Another interesting fact is that the hardest part of our body teeth, softest
and most flexible part tongue and lips and most reflective parts eyes, all find
place in the face. Modeling these parts along with its properties is a great
challenge. Information on human face is scattered in many fields, namely painting,
sculpting, mathematics, anatomy, anthropology, forensic science, and computer
animation. Each field has different concepts, processing style and techniques
to deal with face study. It is important to address all the needs of various
fields, deal with this multidisciplinary study and reconstruct a human face along
with all its details.
To do this Researcher has developed a concept
of AdarshMudra, Jeevak and Vyaktirekha, and has designed and developed a
methodology along with the software system which does not require skilled
personnel to reproduce/reconstruct an image of 3D human face from verbal, photo
or x-ray inputs.
2. Reconstruction Method
Researcher believes that human face can be thought of as a design pattern.
Fig. 1: Bayesian
network for Human Facial feature placement and Face as Design pattern
Each feature is converted into parametric form.
Parameterization is based on
v Anthropometric
Analysis (dimension) and
v Morphological
Analysis (Form and Shape)
Proposed philosophy embarks with Mudra phase in
which generic facial 3D surface is created from AdarshMudra reference sketches.
In PurnaMudra phase, 3D surface then morphed to specific face on the basis of
reference input. Jeevak and Vyaktirekha concepts help to achieve correct facial
proportion with local facial tissue depth. Further facial details can be added
in ShilpaMudra Phase. Mutation of Face can be viewed in BhavMudra Phase. Thus a
complex task of facial reconstruction is accomplished in simple phases.
Fig.2 Phases comprised in Proposed
Research Study
2.1 AdarshMudra
(आदर्श मूर्धन् (Aadarsha muurdhan) – Sanskrit
Word – Ideal Imprint)
AdarshMudra refers to building a template or generic face representing
the Indian population. It confirms norms set by virtuosos and supports
anthropometric dimensions of Indian population.
Fig. 3 Sketch for AdarshMudra
2.2 Jeevak
and Vyaktirekha (जीवक
(Jiivak) – Sanskrit Word – Living, व्यक्तिरेखा (Vyaktirekha) – Sanskrit Word – Features curve
network) Vyaktirekha,
features curve network defined by interconnected facial landmarks. Each
Vyaktirekha point is a landmark point that confines anthropometric dimensions,
anthropometric angles and dimensions. Jeevak is a topology defined by
interconnected skull landmarks working as key points on human skeleton. Jeevak
point may contain the cephalometric information and tissue depth information
associated with related landmark, fat pocket, facial muscle direction and
orientation. There will be corresponding Vyaktirekha point to hold the
relationship. Jeevak point may not present for the facial features composed of
soft tissues and cartilage.
Fig.4 Jeevak and Vyaktirekha Concept
Jeevak and Vyaktirekha will be implemented using bone based method that
uses joint and skinning to connect facial mesh and skull mesh. Eventually
facial mesh can be derived from layered approach comprising lattice
manipulation, forward skeleton kinematics, smooth skin binding and shape
interpolation.
Proposed system attempts to deliver a facial 3D
surface based on verbal explanation and/or front/lateral photographs and/or
Cephalogram or 3D scanned data. Obviously, accuracy of output surface reflects
the correctness and calibration of input.
Proposed system can be viewed as fast facial modeling framework for
modeler in entertainment industry, 3D face visualization tool for a forensic expert,
3D simulated facial surface for surgical planning and quick sketching 3D canvas
to generate face from explanation for police. 3D Facial model can be used be
used as interactive tutor to teach medical students.
3. Experimental Results
4. Future work
Proposed system can be enhanced with more
features listed below:
i.
Skin
texture, can be extracted from the photographs. The skin shader for facial
surface can be developed with various maps to give exact look of an individual.
ii.
Hairstyle
and Haircut can also be added as separate module and will enhance the results.
iii.
Although
current study limits facial variation of Adult Caucasian male of Indian origin,
in future it can be extended to female, wider age group and for other races and
sub races. Detailed morphometric analysis of each feature and its conversion
into various blend shapes will enhance the quality of the system.
iv.
Facial
surface response to dental rearrangement can be explored in future. Orthodontic,
surgical operations may be simulated on 3D face to predict implications.
5. conclusions
The proposed system presents unified path of
understanding the human face from various streams ranging from art to science
and builds a common channel for information exchange. System can work on the
basis of vague verbal explanation or precise input and attempts to deliver the
output. The process is comprised of phase wise development and presents dynamic
and flexible method for facial reconstruction in 3D environment.
References:
1.
Chakrabarti D., Indian Anthropometric Dimensions For Ergonomic design practice,
National Institute of Design, Paldi, Ahmedabad
2.
Chaurasia BD., Human Anatomy Volume Three, Head, Neck and Brain, Third Edition, CBS Publishers and
Distributors, New Delhi, India, 2003, pp 1-30, 36-47, 193-196, 211-212,
215-216, 229-230, 234-235
3.
Clement
JG., Marks MK., Computer Graphic Facial Reconstruction, Elsevier academic press,
USA, 2005, pp 20-23, 35-42, 129-141, 183-195, 221-236
4.
Cordea
MD., Petriu EM., “A 3-D Anthropometric-Muscle-Based Active Appearance Model”,
IEEE Transactions on Instrumentation and Measurement, Vol. 55, No. 1, Feb 2006
5.
DeCarlo,
D. Metaxas D. and Stone M.,“An Anthropometric Face Model using Variational Techniques”,
Department of Computer and Information Science, University of Pennsylvania,
SIGGRAPH ’98.
6.
Fairbanks AT., Fairbanks
EF. Human Proportions for Artists, Washington , USA ,
Fairbanks Art and Books, 2005, pp 101-110, 136-137, 164.
7.
Gould D.,
Complete Maya Programming Volume I, Morgan Kaufmann Publishers, 2005
8.
MacKay J.,
Forensic Art, Lucent Books, Farmington Hills , MI ,
USA , 2009,
pp 8-58
9.
Sahni D., Sanjeev, Singh G., Jit I. , Singh P., Facial soft tissue thickness in northwest
Indian adults Forensic Science International, Volume 176, Issue 2, Pages
137-146,
10.
Taylor
KT., Forensic Art and Illustration,
CRC Press, Boca Raton, Florida, 2000, pp 49-72, 93-106, 341-356, 419-452
11.
Wilkinson
C., Forensic Facial Reconstruction,
University of Manchester, Cambridge University Press, UK, 2008, pp 69-78,
102-119, 124-146, 200-209