CompIMAGE`10 - Computational Biomodeling Lab

International Symposium on Computational Modeling of
Objects Represented in Images: Fundamentals, Methods,
and Applications
Adam’s Mark Hotel, Richardson Room, Buffalo, NY, USA
May 5-7, 2010
Adams’ Mark Hotel, Third Floor
The symposium will be held in Richardson Room.
The banquet will be in Wright Room.
Scientific Program
Tuesday, May 4th
Wednesday, May 5th
Opening Session
Chair: Valentin E. Brimkov
Opening Addresses
Kevin J. Railey, Interim Provost SUNY Buffalo State College
Mark W. Severson, Dean of the School of Natural and Social Sciences
at SUNY Buffalo State College
Opening Talk by Chandrajit L. Bajaj, University of Texas at Austin
Title: Spatially Realistic Multi-scale Modeling from Electron
Coffee Break & Picture
Theoretical Foundations of Image Analysis and Processing
Chair: Peter Balazs
Curvature Estimation for Discrete Curves Based on Auto-adaptive Masks
of Convolution
Christophe Fiorio, Christian Mercat, Frederic Rieux
An Algorithm to Decompose n-Dimensional Rotations into Planar Rotations
Aurelie Richard, Laurent Fuchs, Sylvain Charneau
Tile Pasting Systems for Tessellation and Tiling Patterns
T. Robinson, S. Jebasingh, Atulya K. Nagar, K.G. Subramanian
Collage of Iso-Picture Languages and P Systems
S. Annadurai, V.R. Dare, T. Kalyani, D.G. Thomas
Online Tessellation Automaton Recognizing Various Classes of Convex
H. Geetha, D.G. Thomas, T. Kalyani
Mary Jemima Samuel, V.R. Dare, T. Kalyani
Lunch (on your own) and preparation for the symposium tour
Niagara Falls Tour
Thursday, May 6th
Keynote: Dinggang Shen, University of North Carolina – Chapel Hill
Title: Computational Methods for Quantitative Analysis of Brain
Methods and Applications. Medical Imaging, Bioimaging,
Biometrics, and Imaging in Material Sciences
Chair: Joao Manuel R. S. Tavares
Surface-based Imaging Methods for High-resolution Functional Magnetic
Resonance Imaging
David Ress, Sankari Dhandapani, Sucharit Katyal, Clint Greene, Chandra
Direction-Dependency of a Binary Tomographic Reconstruction Algorithm
Laszlo Varga, Peter Balazs, Antal Nagy
Surface Reconstruction with an Interactive Modification of Point Normals
Taku Itoh
Surface Finish Control in Machining Processes using Haralick Descriptors
and Neuronal Networks
Enrique Alegre, Rocio Alaiz-Rodrıguez, Joaquın Barreiro, Eduardo Fidalgo,
Laura Fernandez
Coffee Break
Theoretical Foundations of Image Analysis and Processing
Chair: Kalman Palagyi
Generalized Perpendicular Bisector and Circumcenter
Marc Rodrıguez, Sere Abdoulaye, Gaelle Largeteau-Skapin, Eric Andres
Digital Stars and Visibility of Digital Objects
Valentin E. Brimkov, Reneta P. Barneva
Omega-arithmetization of Ellipses
Agathe Chollet, Guy Wallet, Eric Andres, Laurent Fuchs, Gaelle LargeteauSkapin, Aurelie Richard
Connectedness of Offset Digitizations in Higher Dimensions
Valentin E. Brimkov
Lunch (on your own)
Keynote: Yongjie Zhang, Carnegie Mellon University, Pittsburg
Title: Image-Based Geometric Modeling and Mesh Generation of
Heterogeneous Domains for Computational Mechanics
Methods and Applications. Medical Imaging, Bioimaging,
Biometrics, and Imaging in Material Sciences
Chair: Renato Natal Jorge
Compact Binary Patterns (CBP) with Multiple Patch Classifiers for Fast
and Accurate Face Recognition
Hieu V. Nguyen, Li Bai
Fast Automatic Microstructural Segmentation of Ferrous Alloy Samples
using Optimum-Path Forest
Joao Paulo Papa, Victor Hugo C. de Albuquerque, Alexandre Xavier
Falcao, Joao Manuel R.S. Tavares
Numerical Simulations of Hypoeutectoid Steels under Loading Conditions,
based on Image Processing and Digital Material Representation
Lukasz Rauch, Lukasz Madej, Bogdan Pawlowski
Customizable Visualization on Demand for Hierarchically Organized
Information in Biochemical Networks
Peter Droste, Eric von Lieres, Wolfgang Wiechert, Katharina Noh
Coffee Break
Image Reconstruction, Computed Tomography, and Other
Chair: David Ress
Graph-Theoretic Image Alignment using Topological Features
Waleed Mohamed, A. Ben Hamza, Khaled Gharaibeh
On the Effects of Normalization in Adaptive MRF Hierarchies
Albert Y. C. Chen, Jason J. Corso
Short communication: The Generalized Orientation Field Transform
Kristian Sandberg
Short communication: Applications of Visual Data Processing to Chemistry
Natalie Nazarenko
Movie “Herbert Hauptman: Portrait of a Laureate”
Wright Room
Dennis K. Ponton, President SUNY Buffalo State College
Dennis L. Hefner, President SUNY Fredonia
Keynote: Venu Govindaraju, University at Buffalo
Title: Biometrics and Security
Friday, May 7th
Keynote: Sargur Srihari, University at Buffalo
Title: Computational Forensics 09:30-10:50
Image Reconstruction, Computed Tomography, and Other Applications
Chair: Khalid Siddiqui
Topology Preserving Parallel Smoothing for 3D Binary Images
Gabor Nemeth, Peter Kardos, Kalman Palagyi
Coding a Simulation Model of the 3D Structure of Paper
Eduardo L. T. Conceicao, Joana M.R. Curto, Rogerio M. S. Simoes,
Antonio A. T.G. Portugal
Crowd Behavior Surveillance Using Bhattacharyya Distance Metric
Md. Haidar Sharif, Sahin Uyaver, Chabane Djeraba
A New Method for Generation of Three-Dimensional Cubes
R. Arumugham, K. Thirusangu, D.G. Thomas
Coffee Break
Special Track on Object Modeling, Algorithms, and Applications
Chair: A. Ben Hamza
Evaluation of New Character Segmentation Approach for Offline Cursive
Handwriting Recognition in the State of Art
Amjad Rehman, Tanzila Saba, Dzulkifli Mohamed, Ghazali Sulong
Non-Linear Least Square Optimization of Intracellular Action Potential
Model Using a Series of Modified Gamma Distribution Functions
GyuTae Kim, Mohammed Ferdjallah
Recognizable Polyhexes Languages and their Acceptors
H. Geetha, J.D. Emerald, D.G. Thomas, and T. Kalyani
Two Metrology Applications in Medical Imaging
Thierry Brouard
Matched Chaotic Maps Watermarking and Authentication
Mohamed Rizk Mohamed Rizk, Said Esmail El Khamy, Amira El Sayed
Using Turning Functions to Refine Shapes
Carlos Frederico de Sa Volotao, Rafael Duarte Coelho dos Santos,
Luciano Vieira Dutra, and Guaraci Jose Erthal
Lunch (on your own) and preparation for walking tour
Buffalo Walking Tour
Keynote Talks
Spatially Realistic Multi-scale Modeling from Electron Microscopy
Wednesday, May 5, 9:00-10:00 AM
Chandrajit L. Bajaj
Center for Computational Visualization,
Department of Computer Sciences,
Institute for Computational Engineering and Sciences,
University of Texas at Austin
Bio-sketch: Chandrajit L. Bajaj is the director of the Center for
Computational Visualization, in the Institute for Computational and
Engineering Sciences (ICES) and a Professor of Computer Sciences at
the University of Texas at Austin. Bajaj holds the Computational
Applied Mathematics Chair in Visualization. He is also an affiliate faculty member of
Mathematics, Electrical Engineering, Bio-Medical Engineering, and also a member of the
Institutes of Cell and Molecular Biology, and Neurosciences, the Center for Learning and
Memory, and the Center for Perceptual Systems. He is an author and editor of over 300
publications, including 225 papers, 25 book chapters, and 1 book and 3 edited volumes. He is on
the editorial boards for the International Journal of Computational Geometry and Applications,
the ACM Transactions on Graphics, the ACM Computing Surveys, the SIAM Journal on
Imaging Sciences, and the International Journal for Computational Vision and Biomechanics. He
is on numerous national and international conference committees, and has served as a scientific
consultant to national labs and industry. He is also a fellow of the American Association for the
Advancement of Science (AAAS) and fellow of the Association of Computing Machinery
Human functional processes are mediated through complicated biochemical and biophysical
interactions amongst proteins, nucleic acids, and other biomolecules. Comprehensive models and
analysis of these interactions, at multiple scales, provide important clues for developing
therapeutic interventions related to infections and disease. In this two part talk I shall first
describe a combination of image processing, and computational geometry algorithms to
efficiently construct adaptive, multi-resolution structure models of target proteins, nucleic acids
that are culpable in the spread of viral infections (e.g. HIV). Next, I shall describe how the multiresolution structure models are utilized to develop a hierarchy of biophysical models of
molecular-molecular (recognition) interaction. Furthermore, I shall describe a fast algorithm
based on non-uniform FFT for estimation of multi-resolution molecular solvation energetics,
while indicating the need for faster computations of multiscale protein binding energetics,
essential for drug screening and discovery.
This is joint work with members of UT- computational visualization center, and ICES, as
well as collaborators at UCSD, the National Cancer Institute, and the Scripps Research Institute.
Biometrics and Security
Thursday, May 6, 8:30-9:30 PM
Venu Govindaraju
Department of Computer Science and Engineering
University at Buffalo
Biosketch: Prof. Govindaraju is a UB Distinguished Professor of
Computer Science and Engineering at State University of New
York, Buffalo. He received his B-Tech (Honors) from the Indian
Institute of Technology, Kharagpur, and his PhD. from SUNYBuffalo. Prof. Govindaraju has authored more than 300 scientific
papers and supervised the dissertation of 20 doctoral students. His
seminal work in handwriting recognition was at the core of the first handwritten address
interpretation system used by the US Postal Service. Prof. Govindaraju is the founding director
of the Center for Unified Biometrics and Sensors. He has won several awards for his scholarship,
including the ICDAR Young Investigator Award (2001), the MIT Global Technovator Award
(2004), the HP Open Innovation Award (2008, 2009), and the IEEE Technical Achievement
Award (2010) . He is a fellow of the IEEE, ACM, IAPR.
Transforming raw biometric data pertaining to the identity of human subjects (face, voice, gait,
etc.) into a form that is suitable for information retrieval remains a challenging open problem,
spanning many research areas including video and audio processing, computer vision, spatiotemporal reasoning, and information retrieval. Towards this end, we have developed a new
paradigm called evolutionary identification. That is, the evidence of identity of individuals
accrues, or evolves, over the course of events as they get captured on various biometric devices
at different locations. This is in contrast to object recognition paradigms to date where the input
signal must be classified (even if only a soft decision is made) immediately upon its acquisition.
Furthermore, unlike earlier approaches to the problem of tracking people, we will not require a
complete coverage of the monitored space by massive numbers of sensing devices; rather, we
explore the more realistic scenario where biometric capture devices are placed only at certain
zones, such as hallways, rooms, entrances, etc. The talk will conclude with a brief overview of
the challenges with biometric systems that must be met before it gains broad based citizen
Computational Methods for Quantitative Analysis of Brain Diseases
Thursday, May 6, 8:30-9:30 AM
Prof. Dinggang Shen
Department of Radiology, BRIC, and Computer Science
University of North Carolina - Chapel Hill, USA
Bio-sketch: Prof. Shen received all of his degrees from Shanghai Jiao
Tong University. Before joining UNC-CH, he worked as a faculty
member in the University of Pennsylvanian and the Johns Hopkins
University. His research interests include medical image analysis,
computer vision, and patter recognition. With his colleagues, Prof.
Shen has developed many innovative and practical methods for
deformable segmentation (AFDM), registration (HAMMER, CLASSIC, ORBIT, RABBIT,
TIMER), and neuroimage classification (COMPARE and STEP). These methods have been
applied for diagnosis of brain diseases (e.g., AD, MCI, and schizophrenia), cardiac disease,
prostate cancer, and breast cancer. He has published over 200 papers in the international journals
and conference proceedings. Currently, he is a director for the Image Display, Enhancement, and
Analysis (IDEA) Lab in the Department of Radiology, and also a director of medical image
analysis core in the Biomedical Research Imaging Center (BRIC) at UNC-CH.
This talk will summarize our work on analysis of MR brain images. Our main research goal is to
develop automated image analysis methods for precisely quantifying subtle and complex
structural/ functional changes in the brains, to be used for early detection of brain diseases, such
as Alzheimer's Disease (AD). Accordingly, we developed a 3D brain registration method, called
HAMMER, which has been successfully applied to many large clinical research studies and
clinical trials involving more than 8,000 MR brain images. In order to measure the tiny
longitudinal brain changes, i.e., due to AD, we also have developed a 4D (3 spatial dimensions +
1 temporal dimension) brain registration algorithm and obtained more accurate measurement
results, compared to those by 3D registration algorithm. In addition, we have developed
multivariate analysis methods, based on support vector machine, to jointly consider all
structural/functional changes for determining the group difference between brains, due to
diseases, aging, or development. This method has been used for classifying schizophrenia
patients from normal controls, and for lie detection based on the functional MR images. Details
of these 3D and 4D registration algorithms and nonlinear brain analysis methods will be
discussed in this talk. Some new applications, i.e., in early brain development from two weeks to
1 year old and 2 year old, will be also presented.
Computational Forensics
Friday, May 7, 8:30-9:30 AM
Prof. Sargur (Hari) N. Srihari
Department of Computer Science and Engineering
University at Buffalo
Bio-sketch: Prof. Sargur Srihari received a B.Sc. in Physics and
Mathematics from the Bangalore University (National College) in 1967,
a B.E. in Electrical Communication Engineering from the Indian
Institute of Science, Bangalore in 1970, and a Ph.D. in Computer and
Information Science from the Ohio State University, Columbus in 1976.
Prof. Srihari is a SUNY Distinguished Professor in the Department
of Computer Science and Engineering at the University at Buffalo, The State University of New
York. With support from the United States Postal Service for over 20 years, he founded CEDAR,
the Center of Excellence for Document Analysis and Recognition, in 1991, which had a major
impact on the development of various aspects of the field. Research at CEDAR led to a new
thread of work leading to the first large-scale handwritten address interpretation systems
deployed by the IRS and by the USPS.
Prof. Srihari's handwriting recognition work led to the first handwritten address interpretation
system ever used by post offices in the world. The software developed by his research team was
deployed on a national scale by the United States Postal Service, which was later extended to
UK-Royal Mail and Australia Post.
Prof. Srihari's work on computational forensics has had an impact both on the courts and on
the software tools used by forensic scientists. His studies on individuality measurement is widely
cited in the context of forensic testimony. His work also led to a software system in use by
forensic document examiners worldwide. His most recent work is on probabilistic
characterization of fingerprint evidence.
Prof. Srihari is an author of over 300 research papers, of which 65 are in journals (including
one in the first issue of IEEE Transactions on Pattern Analysis and Machine Intelligence) and 6
patents. He has edited three books, served as principal advisor to 34 doctoral students and served
as general chair of several conferences/workshops, including a recent one defining the field of
computational forensics.
Prof. Srihari's honors include: Fellow of the Institute of Electronics and Telecommunications
Engineers (IETE, India) in 1992, Fellow of the Institute of Electrical and Electronics Engineers
(IEEE) in 1995, Fellow of the International Association for Pattern Recognition in 1996 and
distinguished alumnus of the Ohio State University College of Engineering in 1999.
Forensic analysis has as its objective whether observed evidence arises from the same source as
of a known. Computational forensics is analogous to similar efforts in other scientific
disciplines, e.g., computational geometry, computational vision, computational biology,
computational chemistry, etc., where human-based approaches to convert data to knowledge are
translated into algorithms and software. Computational forensics can play a role in overcoming
several shortcomings of the forensic sciences which have received much recent public attention
and criticism. The presentation gives an ontology of forensics distinguishing the terms digital
forensics, classical forensics and computational forensics. Three main research topics of
computational forensics which span several sub-disciplines are: (i) the individuality problem,
which is the quantification of the “degree” of uniqueness provided by a modality or evidence, (ii)
the search problem, which is to narrow-down possibilities in a database using evidence as query,
and (iii) the comparison problem, which involves determining the degree of match between the
evidence and known while considering possible variability within the known. Statistical
approaches to solving each of these problems will be described. The solutions are illustrated
with examples from DNA, finger-prints, shoe-prints, handwriting and signatures.
Image-Based Geometric Modeling and Mesh Generation of Heterogeneous
Domains for Computational Mechanics Thursday, May 6, 1:30-2:30 PM
Prof. Yongjie (Jessica) Zhang
Director of Computational Biomodeling Laboratory
Department of Mechanical Engineering
Carnegie Mellon University, USA
Bio-sketch: Prof. Zhang received her B.Eng. in Automotive
Engineering (1996) and M.Eng. in Engineering Mechanics (1999), all
from Tsinghua University, China; M.Eng. in Aerospace Engineering
and Engineering Mechanics (2002) and Ph.D. in Computational
Engineering and Sciences (2005) from the University of Texas at
Austin. She is the director of Computational Biomodeling Laboratory at Carnegie Mellon
University. Her research interests include Computational Geometry, Mesh Generation, Computer
Graphics, Visualization, Finite Element Method, Isogeometric Analysis and their applications in
Computational Biomedicine, Computational Biology and Engineering. Prof. Zhang has
developed many novel meshing techniques for quality 2D and 3D finite element mesh
generation, which have been used in a lot of applications at various scales. She has published
about 50 papers in the international journals and conference proceedings.
With finite element method and scanning technology seeing increased use in active research
areas such as biomechanics, there is an emerging need for quality mesh generation of the
spatially realistic domains that are being studied. In images obtained from various scanning
techniques like CT/MRI, the domain of focus often possesses heterogeneous materials and/or
functionally different properties. For example, the MRI brain data is segmented into 48 subareas, with each colored area demarked as possessing specific characteristic functionality. In
finite element analysis, these heterogeneous materials are grouped into separate material regions
with individual physical/chemical attributes or material coefficients. For each of the partitioned
material regions, high fidelity geometric models and quality meshes are needed, with meshes
conforming at the material boundaries. Although there have been tremendous progresses in the
area of surface reconstruction and 3D geometric modeling, it still remains a challenge to generate
desirable models for such complicated domains.
I will present details of meshing pipelines, especially octree-based algorithms to extract
adaptive and quality 2D (triangular or quadrilateral) and 3D (tetrahedral or hexahedral) meshes
of volumetric domains, conforming to boundaries defined as level sets of a scalar function on the
domain. Guaranteed-quality all-quad meshing, feature preservation, and automatic meshing for
multi-material domains will be discussed. Besides piecewise linear element meshes, a skeletonbased sweeping method is developed to construct hexahedral solid NURBS for blood vessels
from imaging data, then a wavelets-based scheme is used to simplify and fair the NURBS
surface with continuity preservation, especially at the interface shared by multiple patches. The
constructed solid NURBS have been successfully used in isogeometric analysis of blood flow. In
this talk, I will additionally present two main applications of our meshing schemes: patientspecific geometric modeling from CT/MRI data, and implicit solvation models of biomolecular
structures for multi-scale models for the Neuro-Muscular Junction synaptic system at both
molecular and cellular scales.