IIT's Visual Computing Lab
The Visual Computing Lab is a research group within the Computer
Science department which is
directed by Dr. Gady Agam. The mission of the lab is to conduct
in advanced computational techniques concerning Visual Computing.
Visual Computing is a broad term that encompasses several areas in
computer science including: computer vision, computer graphics, image
and video processing, geometric modeling, and pattern recognition.
In general, we target the development and application of image
techniques to medical, scientific, industrial, and military problems in
which automatic or semi-manual processing of images and volumetric data
is required to inspect, recognize, enhance, or complement other sensory
information. Applications of image based rendering to augmented and
virtualized reality and other synthetic environments. Image based
perceptual user interfaces, multimedia databases and multimedia
computing, image based biometrics, document analysis and recognition.
Research funding to projects involving the Visual Computing Lab in
recent years totaled
roughly $1.5M. The majority of the funding was provided by the National
Science Foundation (NSF), the National Institute of Health (NIH), the
National Security Agency (NSA), and the
Pritzker Institute for Biomedical Engineering.
keywords: 3D Modeling, Artificial Intelligence, Bioinformatics,
Biomedical Engineering, Classification Theory, Computer Graphics,
Computer Storage and Retrieval, Computer Vision, Diagnostic Imaging,
Digital Imaging, Digital Systems Processing, Geometry, Image Analysis,
Image Processing, Image Reconstruction and Restoration, Image Science,
Machine Vision, Medical Imaging, Medical Informatics, Multimedia or
Interactive Communications Technology, Pattern Recognition, Robotics,
Signal Analysis, Signal Processing, Surveillance Systems, Target
Recognition, Virtual Reality
Current research activities in the lab focus on the following areas of
applications: medical imaging, document imaging, geometric modeling and
3D biometrics. A short description of recent and ongoing activities in
the lab is provided below.
Research on medical imaging is done in conjunction with researchers in
the radiology department at the university of Chicago and the
engineering department at IIT.
Research on document imaging is done in conjunction with researchers in
the University of Maryland, the university at Buffalo, and the IR group
Some of the research on geometric modeling is done in conjunction with
researchers in the mechanical engineering department at IIT.
Below is a short review of examples application of current and past
| Diffusion tensor warping.
Visual analysis of 3D neuroimaging modalities such as fMRI, MRI and DTI
brain imaging is a significant component in research and clinical
practice. Using imaging techniques it is possible to assess brain
activity in humans and study the reorganization of brain activity after
disabling conditions. This can then be used to asses conditions, plan
interventions, and monitor recovery.
We are developing computational techniques for automated analysis,
interpretation, and visualization of brain structure and function in
fMRI, MRI and DTI brain imaging data. We are particularly interested in
developing computational tools to facilitate the study of functional
morphometric variability in human brains by registering brain data sets
of different subjects with and without abnormalities.
| Automated lesion detection.
Lesion segmentation in MRI scans is used for lesion quantification as
pertaining to various medical conditions. We developed a novel
for chronic stroke lesion segmentation based on multiple
modalities including T1-weighted and T2-weighted images as well as
diffusion tensor-based modalities.
The approach is based on a mixture-parametric probabilistic model
the model parameters are optimized by maximizing the incomplete-data
log-likelihood function through expectation maximization. The
mixture components are selected to have Cauchy distributions.
A probabilistic prior is computed by evaluating the feature vectors for
a set of registered brain scans in a control set.
| Vasculature extraction in
Vessel tree reconstruction in volumetric data is a necessary
prerequisite in various medical imaging applications. We are developing
a novel approach to vessel tree reconstruction and its
application to nodule detection in thoracic CT scans using
correlation-based and probabilistic enhancement filters.
The correlation-based enhancement filters we are developing depend on
first-order partial derivatives and so are less sensitive to noise
compared with Hessian-based filters. Additionally, multiple sets of
eigenvalues are used so that a distinction between nodules and vessel
junctions becomes possible.
| Vessel enhancement.
Vessel enhancement in volumetric data is a necessary prerequisite
in various medical imaging applications.
Ideally, vessel enhancement filters should enhance vessels and vessel
junctions while suppressing nodules and other non-vessel elements. We
developed probabilistic vessel models from which novel
vessel enhancement filters capable of enhancing junctions while
suppressing nodules are derived. The filters we developed are based on
eigenvalue analysis of the structure tensor which is a first order
differential quantity and so are less sensitive to noise.
| Complex document information
We investigate the problem of content-based image retrieval in the
of complex document images. Complex documents typically start out on
paper and are then electronically scanned. These
documents have rich internal structure and might only be available in
image form. Additionally, they may have been produced by a combination
of printing technologies (or by handwriting); and include diagrams,
graphics, tables and other non-textual elements. Large collections of
such complex documents are commonly found in legal and security
We have developed a test collection of several million documents and a
prototype system to ingest it.
| Degraded document
There are numerous collections of scanned documents.
of such collections include historical collections, legal depositories,
medical archives, and business archives. Moreover, in many situations
such as legal litigation and security investigations scanned
are being used to facilitate systematic exploration of the data. It is
almost always the case that scanned documents suffer from some form of
degradation. Large degradations make documents hard to read and
substantially deteriorate the performance of automated document
processing systems. We are developing statistical techniques for
degraded document image
enhancement in this context.
| Writer identification.
Writer identification in offline handwritten documents is a difficult
task with multiple applications such as authentication, identification,
and clustering in document collections. For example, in the context of
content-based document image retrieval, given a document with
handwritten annotations it is possible to determine whether the
were added by a specific individual and find other documents annotated
by the same person. We are investigating techniques for deriving
canonical stroke frequency descriptors from handwritten text to
writers. We demonstrated that a relatively small set of canonical
strokes can be
successfully employed for generating discriminative frequency
| Signature recognition.
Offline signature recognition is an important form of biometric
identification that can be used for various purposes.
Similar to other biometric measures, signatures have inherent
variability and so pose a difficult recognition problem.
We are developing a novel approach for solving the curve
correspondence problem that is not limited by the requirement of one
dimensional parametrization. Our approach utilizes particle
dynamics and minimizes a cost function through an iterative solution of
a system of first order ordinary differential equations. The
approach is, therefore, capable of handling complex curves for which a
simple parametrization is not available.
| Adaptive mesh subdivision.
We are developing novel techniques to adaptively subdivide mesh
using existing (e.g. , Loop, and Catmull-Clark)
subdivision schemes. In
modifying global adaptive subdivision schemes to perform local
subdivision, it is necessary to guarantee smooth transition between
subdivided regions and regions left at the original level to
prevent the formation of surface artifacts at the boundaries between
such regions. We are developing techniques for incremental adaptive
subdivision that prevent and decouple cracks and degenerate faces by
subdividing iteratively without forming artifacts. User selectable
measures enhanced by hysteresis thresholding, guide a dynamic selection
Smoothness across different subdivision depths is obtained by the
postponement of local atomic operations.
| Accurate curvature
Accurate curvature estimation in discrete surfaces is an important
problem with numerous applications. Curvature is an indicator of ridges
and can be used in applications such as shape analysis and recognition,
object segmentation, adaptive smoothing, anisotropic fairing of
irregular meshes, and anisotropic texture mapping. We are developing a
new framework for accurate curvature estimation in
discrete surfaces. The proposed framework is based on local
directional curve sampling of the surface where the sampling frequency
can be controlled. This local model has a large number of degrees of
freedoms compared with known techniques and so can better represent the
| 3D face recognition.
Automated 3D object analysis and recognition is an essential component
in numerous applications such as manufacturing automation, quality
control, autonomous navigation, indexing and retrieval of object
computer aided diagnosis in medical imaging, and biometrics. Biometrics
in particular has become a focal point of research with significant
importance. We are investigating a general framework that may be
to a wide range of applications, and target specifically the
of 3D surface-based biometrics.
| Novel view-synthesis.
Synthesizing new views from existing ones has numerous
applications. Approaches for view synthesis rely on dense or
sparse matching between existing views. In all cases, some parts of the
original images are inevitably unmatched and so the synthesis
of new views of the corresponding regions need to rely upon sparse sets
of matched points. Triangulation based upon sparsely matched points
constitutes a possible solution for the handling of unmatched
regions. However, such a triangulation should respect the underlying
geometry of the corresponding scene.
We developed techniques for physically valid scene triangulation which
is then used for novel view synthesis.
Duc Ngoc Tran