ATTENTION: THIS WEB SITE HAS MOVED.
The pages you are looking at are no longer being maintained.
Please go to http://www.poly.edu/cis/
to visit the new site of the Department of Computer and Information Science
at Polytechnic University.
Visual Information Processing
(Profs. Chiang,
Gluckman,
Memon,
Wong)
Another major research area in the department is concerned with the
analysis, storage, transmission, and synthesis of images, video,
and multimedia data. A group of faculty is working on a variety of
problems in this domain, including Computer vision and image analysis,
image and video compression and transmission, watermarking &
protection mechanisms, computer visualization,
and computer graphics. Work is supported by multiple grants, and also
involves cooperation with several faculty in the Electrical Engineering
department.
Image/Video Analysis & Pattern Recognition:
Prof. Wong has worked extensively on the analysis of image and video
data. A lot of his current research focuses on the analysis of
document images, such as maps or engineering drawlings, and on
information retrieval from images and video. For example, his work
has resultied in a number of novel techniques and algorithms for
noise filtering, segmentation, and lossy compression of engineering
drawings. The segmentation process separates an engineering drawing
into separate layers, which can then be compressed, retrieved, and
analyzed separately for best overall performance. Prof. Wong's lossy
compression algorithm, based on straight-line-extraction and
higher-order-context-modeling techniques, allow engineering drawings
to be compressed down to about 1% of its original size without loss
of visual quality or useful information.
Prof. Wong has also developed novel techniques to retrieve images
from image/video databases. With the rapid advances in digital
technology, more and more databases are multimedia in nature,
containing images and video in addition to textual information.
Currently, most video databases are manually indexed based on
textual annotations in an often tedious and time consuming process.
Using a computerized approach, indexing and retrieval are performed
based on features extracted directly from the video. One novel
feature he developed was the Augmented Histogram where spatial
information is added to a color histogram of an image, improving
retrieval precision and recall over conventional histograms.
Computer Vision:
Advances in both image sensors and the processing of image data will
lead to devices that can intelligently gather information about the world
and create digital models that mimic the real world in their complexity.
These digital models will be used for scientific visualization, education,
robotic exploration, and historical preservation.
The goal of Prof. Gluckman's research is to develop new imaging technologies
and computational techniques for accurately sensing important scene
properties such as shape, motion and reflectance. In particular, we are
interested in designing stereo vision sensors and camera motion sensors.
The stereo sensors he is developing have led to real-time systems for
capturing three dimensional information. By altering the distribution of
viewing rays, new cameras are being designed that simplify the estimation
of camera motion. When coupled with new processing techniques, these
devices will be able to rapidly and accurately measure the motion and
shape of a moving object.
Multimedia Compression: Another active area of research is
concerned with compression techniques for various forms of image
and multimedia data. One faculty member, Prof. Memon, was involved
in an international collaboration for designing a novel lossless
compression scheme in response to the JPEG (Joint Pictures Expert
Group) committee of the International Standards Organization's (ISO)
call for proposals for a new standard. His scheme, called
CALLIC ranked first in the evaluations conducted by ISO, and
significantly influenced the adopted standard. Other work in this
area includes compression techniques for volumetric data sets, by
Profs. Chiang and Memon, document compression research by Prof. Wong
(see above), and work on compressing HTML data and web graph structure
by Profs. Memon and Suel.
Multimedia Content Protection: Research in this area focuses
on techniques for preventing or detecting unauthorized use, copying,
or alteration of multimedia content, such as images, audio, or video.
One basic approach is based on Digital Watermarking, where
a usually imperceptible signal is added that identifies the source,
establishes ownership, or detects alteration of the data. In his work,
Prof. Memon has demonstrated several serious weaknesses in current
watermarking techniques, and has developed new techniques that
overcome these shortcomings. His work on secure distribution of
multi-media content is performed in collaboration with researchers
in Intel, Hewlett-Packard and Panasonic. Other related work in the
department concerns watermarking of documents (Prof. Wong) and of
software (Profs. Memon and Naumovich).
Computer Graphics & Visualization: Work in this area is
concerned with the synthesis of realistic images, and the visualization
of highly complex data sets such as those obtained from scientific
simulations and measurements and from medical applications.
Recent advances in three-dimensional acquisition, simulation,
modeling and virtual-reality techniques have led to massive datasets
that exceed the main memory size and the interactive rendering
capability of current graphics hardware. As the complexity of
graphics datasets increases, I/O-efficiency becomes more and more
important, but very few of the existing techniques
explicitly consider this issue. Isosurface extraction is one of the
most effective techniques for visualizing and studying volumetric
data sets where objects are given by 3D sample points over their
volume. This techniques allows us to visualize the interior of the
objects and study them in detail; however, the data sets involved
tend to be huge, making I/O-efficient techniques extremely important.
Prof. Chiang works on developing isosurface techniques based on
I/O-optimal indexing structures and advanced partitioning methods
that achieve a speed-up of one to two orders of magnitude in query
time for large data sets, using disk space of only 1.1-1.5 times
the original data size.
Prof. Chiang also developed the first conflict prediction code based
on geometric hashing to support free flight in Air Traffic Control,
which is now being used by Seagull Technology in a NASA project. In
addition, he devised the first out-of-core techniques that can
efficiently perform both progressive-mesh simplification and
view-dependent rendering for polygonal models larger than main memory,
and received the Best Paper Award in Eurographics 2000.
