Mr. Barry Barnett
IBM RS6000 RAS Engineer
Austin, TX
Wednesday, April 26th, 9:00 AM, ENS 602
Video encoding is an important enabling technology for current and future digital video applications. The design of the video encoder determines many, if not all, of the performance characteristics of the video application. Good video encoder design must consider a number of issues that vary according to the requirements of the specific application and/or end user. These issues include encoder rate versus distortion, coding complexity, encoding versus decoding symmetry, variable versus constant bitrate encoding, transmission channel bandwidth and performance characteristics, and compatibility with accepted standard video formats and encoders like MPEG-1, MPEG-2, MPEG-4, H.261 and H.263.
Motion Compensated Visual Pattern Image Sequence Coding (MCVPISC) is a
video encoding model introduced here that was uniquely designed to meet the
large range of performance criteria encountered in the development of
digital video applications. The MCVPISC encoder design is based on a model
of the Human Visual System (HVS) that can accommodate these design
requirements. It makes use of a visual pattern codebook that represents the
most visually significant information for a given fidelity criterion, and
incorporates a motion compensation technique that is based on scalable
predictive differential encoding. MCVPISC performance results on the MPEG-2
test sequences Football and Tempete indicate excellent coding efficiency
and complexity, as well as visual quality. MCVPISC is also demonstrated to
be compatible with Discrete Cosine Transform (DCT) based video encoders
though a technique named the Linear Complexity VPIC-to-DCT Transform.
Videoconferencing is an important and highly demanding digital video
application that was chosen to prototype and demonstrate the capabilities
of the MCVPISC software encoding paradigm. In order to express the
performance capabilities of the computer videoconferencing system video
encoder, the number of full-motion (30 frames/sec) multiparty
videoconferencing sessions was chosen as the figure of merit. This metric
was analytically modeled in order to create a distinctly new design
methodology for videoconferencing system development. Videoconferencing
prototype systems for the monochrome and color versions of the MCVPISC
encoder were developed. Performance measurements indicate that the MCVPISC
encoder can concurrently support multiple full-motion multiparty
videoconferencing sessions. The measurements are also demonstrated to be in
close agreement with the performance predicted by the model.
Mr. Barnett has been employed by IBM since 1984. He has worked on the design and development of IBM enterprise disk servers, RS/6000 servers, Fibre Channel storage servers, high availability clustering software for AIX, security solutions for information technologies, and most recently, Storage Area Network technologies. He received the IBM Outstanding Technical Achievement Award for his contributions to the IBM 3390 program in 1990, and has recently filed two U.S. patents in the field of Storage Area Networks.
A list of Telecommunications and Signal Processing Seminars is available at from the ECE department Web pages under "Seminars". The Web address for the Telecommunications and Signal Processing Seminars is http://anchovy.ece.utexas.edu/seminars