|
 Visual Computing
Group
Visual
Computing Group is a research group under Media and Communication Lab. Our
research fields are shape analysis and data mining. Current research
interests include shape analysis, object recognition, geometric modeling,
and 3D model retrieval. Our group has established cooperative relations
with many other research groups, such as Dept. of Neurology, University of
California, Los Angeles, CIS Dept. of Temple University, Microsoft Research
Asia and Lotus Hill Institute.
Shape Analysis:
1) Shape Representation
Skeleton,
or called Medial Axis,
is a very popular shape descriptor for shape representation and
recognition. The drawback of skeleton is its sensitivity to boundary noises
and deformations: little noise or a variation of the boundary often
generates redundant skeleton branches that may seriously disturb the
topology of the skeleton’s graph. As shown in Fig (a), the skeleton of the
horse is too grassy because of the noises and the short corners on the
contour. Thus, skeleton pruning is usually an essential way to obtain the
visual skeletons required by skeleton-based shape matching. Fig (b) shows the
visual skeleton of a horse, which is the pruned skeleton of (a) with the
proposed pruning method.
(a) (b)
2) Shape Matching and Classification
Skeleton-based
shape matching often has two difficulties: 1) Some skeletons are too grassy
to be matched; 2) Graph matching is an NPC problem, so the complexity of
the skeleton matching algorithm is usually too high. At the beginning, all
the skeletons have been pruned with our pruning method. The main idea is to
match skeleton graphs by comparing the geodesic paths between skeleton
endpoints. In this way, skeleton graph can be transferred into vector
sequence. Thus, the step for skeleton matching can be transferred into the
step for sequence matching.
3) Visual Curvature
Visual
curvature is a new definition of curvature; it can apply to regular curves
as defined in differential geometry as well as to turn angles of polygonal
curves. In this way, we unify the definition of curvature on both regular
curves and polygonal curves.
Visual curvature is
multi-scale; it can ignore small details when calculating the curvature at
a relative larger scale. Thus, it is suitable for many visual processing,
such as curve evolution and salient point detection.
Curve Evolution
of a noisy pentagra
Detected Corners of a noisy plane
4) Data Mining & Machine Learning
There exist
many data mining techniques recently. Our team aims at implementing these
techniques as tools in the computer vision, which help us find the useful
and potential information for low, middle and high level vision problems
such as segmentation and object recognition. Another research topic focuses
on optimization of the classification and clustering models. One
interesting and challenging work of our team is committed to find the
optimal model selection scheme for these models and analyze the influence
of the unbalance training data.
Some
documents in related domain must be read and some skills are required in
our group.
Related Domain:
Computer Graphics, Computer Vision, Pattern
Recognition, Digital Image Processing, Machine Learning, Data Mining,
Geometry Modeling and etc.
Skill
required:
Mathematics: Basic Statistics Theory, Probability and
Random Process, Functional Analysis, Manifold, Matrix Theory, Differential
Geometry, Integral Geometry.
Programming: Qualified programming skill on C and C++,
basic programming skill for 3D models (OpenGL, or Direct3D), familiar with Matlab and Latex. Java or C# is also encouraged.
English: Excellent reading and writing, good speaking
English (recommended).
Publications:
[1].Quannan Li, Yu Zheng, Xing Xie, Yukun Chen, Wenyu Liu,
Wei-Ying Ma, "Mining User Similarity Based on Location History",
ACM GIS 2008.PDFNEW
[2].Xiang Bai, Xingwei Yang, L.J. Latecki,
and Wenyu Liu, "Computing Stable Skeletons
with Particle Filters", the 10th Pacific Rim Int. Conf. on Artificial Intell. (PRICAI), Hanoi, Vietnam, 2008. (AR = 21.3% for
Oral Presentations).PDF
[3].Quan-Nan Li, Xiang Bai, and Wen-Yu Liu, "Skeletonization
of Gray-Scale Images from Incomplete Boundaries", IEEE International
Conference on Image Processing (ICIP'08), San Diego, USA, 2008.PDF
[4].Xingwei Yang, Xiang Bai, Longin Jan Latecki, Zhuowen Tu, "Improving Shape Retrieval by Learning Graph
Transduction "：(Best Ever
Retrieval Rate of 91% on MPEG-7!) , ECCV, to appear in 2008.PDF
[5]. Hairong Liu, Longin J Latecki, and Wenyu Liu, "A Unified Curvature Definition for
Regular, polygonal, and Digital Planar Curves",Editorial
Manager(tm) for International Journal of Computer Vision PDF
[6]. Xiang Bai, Xingwei Yang, and Longin Jan Latecki, " Detection and
Recognition of Contour Parts Based on Shape Similarity", Pattern
Recognition (PR), to appear in 2008. PDF
[7]. Xiang Bai, Longin Jan Latecki, and Wenyu Liu, “Skeleton Pruning by Contour Partitioning
with Discrete Curve Evolution”, IEEE
Trans. Pattern Analysis and Machine Intelligence (PAMI), 29(3):
449-462, 2007. PDF
Matlab code for skeleton pruning:
http://www.cis.temple.edu/~latecki/Programs/skeletonPruning07.htm
[8]. Xiang Bai, and Longin Jan Latecki, “Path
Similarity Skeleton Graph Matching”, IEEE
Trans. Pattern Analysis and Machine Intelligence (PAMI), to appear. PDF
[9]. HaiRong Liu, Longin Jan Latecki, Wenyu Liu, Xiang Bai. Visual
Curvature, IEEE
Computer Society Conference on Computer Vision and Pattern Recognition
2007, CVPR 2007. PDF
[10]. Xiang Bai, and Longin Jan Latecki, “Discrete
Skeleton Evolution”, International
Conference on Energy Minimization Methods in Computer Vision and Pattern
Recognition (EMMCVPR),
Aug. 2007, oral presentation (AR
= 15% for Oral Presentations). PDF
[11]. Longin Jan Latecki, Quan-Nan Li, Xiang Bai, and Wenyu Liu, “Skeletonization Using SSM of the Distance Transform”, IEEE International Conference on Image
Processing (ICIP),
Sep. 2007. (AR = 45%)
PDF
[12]. Xiang Bai, Longin Jan Latecki, and Wenyu Liu, “Skeleton Pruning by Contour
Partition”, 13th
International Conference on Discrete Geometry for Computer Imagery (DGCI), pp: 567-579, October 2006. (AR = 28% for Oral Presentations)
PDF
[13]. Wenyu Liu, Xiang Bai, and Guang-Xi Zhu, “A
skeleton-growing Algorithm Based on Boundary Curve Evolution”, ACTA AUTOMATICA SINICA,
32(2): 255-262, 2006. (In Chinese)
[14]. N. Adluru, L.J. Latecki, R. Lakaemper, T.
Young, X. Bai, and A. Gross. Contour Grouping
Based on Local Symmetry. 11th
IEEE Int. Conf. on Computer Vision (ICCV), Rio de Janeiro, Brazil, October
2007. PDF
[15]. Xiang Bai, Xing-Wei
Yang, and Longin Jan Latecki,
“Shape Classification Based on Skeletons”, International Journal of Pattern Recognition and
Artificial Intelligence (IJPRAI),
to appear.
[16]. Xing-Wei Yang, Xiang Bai,
and Longin Jan Latecki,
“Shape Classification Based on Skeleton Path Similarity”, International Conference on
Energy Minimization Methods in Computer Vision and Pattern Recognition
(EMMCVPR),
Aug. 2007, oral presentation (AR
= 15% for Oral Presentations).PDF
[17]. Quan Zhou, Wenyu Liu, Cunhui Chui, A
Novel Macroblock Layer Rate Control with Low Bit
Rate, Journal of Image and Graphic, 9(8), 2007.
[18]. Xi Geng, Wenyu Liu, Hairong Liu, Force
Field Based Expression for 3D Shape Retrieval, 12th International
Conference on Human-Computer Interaction, 2007, (LNCS). PDF
Useful
Link
Instructions of Publications
Computer Vision Homepage @CMU
http://www.cs.cmu.edu/~cil/vision.html
Prof. Dr. Longin Jan Latecki @Temple University
http://www.cis.temple.edu/~latecki/
Prof. Dr. Song Chun Zhu @UCLA
http://www.stat.ucla.edu/~sczhu/
Prof. Dr. Zhuowen Tu @UCLA
http://www.loni.ucla.edu/~ztu/index.html
Prof. Dr. Alan Yuille @UCLA
http://www.stat.ucla.edu/~yuille/
Dr. Lei Zhang @MSRA
http://research.microsoft.com/users/leizhang/
|
Members
There
are 7 members in our current group:
Current
PhD students：
Team
Leader: Xiang
Bai (2005 Supervisor: Wenyu Liu, Zhuowen Tu and L.J. Latecki)
Microsoft Fellowship 2007, Joint PhD student at UCLA supervised by Zhuowen Tu
Shape Representation, Shape Matching and Classification, Object Detection
and Contour Grouping
Email: xiang.bai@gmail.com
Xingwei Yang (2006 Supervisor: L.J. Latecki)
Shape Matching and Classification)
Email: xingwei.yang@temple.edu
Jun
Chen (2006 Supervisor: Wenyu
Liu)
Shape Matching and Classification)
Email: chenjun71983@163.com
Current
MSc students：
Yu
Zhou 2009
Shape Analysis, Machine Learning
Email: zhouyu.hust@gmail.com
Xinggang Wang 2009
Object Detection and Recognition
Email: wxghust@gmail.com
Current
undergraduate students：
Bo
Wang
Qiao Zhang
Excellent
previous students:
Former
PhD students：
Hairong Liu (Now: Postdoc fellow at NUS, Singapore)
Former
MSc students：
Yuhua Zheng (Now: PhD candidate at Stevens Institute of Technology, USA)
Juntao Liu(Now: Faculty member of Mechanical Engineering Institute,
Shijiazhuang)
Quanan Li (Now: PhD candidate at UCLA, USA)
Chengqian Wu (Now: Master student at NCSU, USA)
Yan
Zhou (Now: Marketing Executive of
P&G)
Xi Geng(Now:
Management Trainee of Cisco Co.Ltd)
Yi Peng(Now:
Researcher of Briontech CO.Ltd)
Former
undergraduate students:
Xiang
Huang (Now: PhD candidate at
Northwestern University, USA)
Xingwei Yang (Now: PhD candidate at Temple University, USA. He is still the
member of our group)
Xiaojun Yang
Yifan Li (Now: PhD candidate at NTU, Singapore)
Jian Zhou (Now: PhD candidate at NUS, Singapore)
Photos
:)
Hairong Liu
Xiang Bai
Quannan Li
|
