Research Summaries and Recent Results
We have included here a brief description of some of the research projects
currently underway within SIPI - clicking on the titles will connect you
to a more detailed description. This list is representative rather than
comprehensive and will be expanded in the near future. Detailed descriptions
of this work are available in the open literature, more recent results are
disseminated through the SIPI
Technical Report Series.
 : ( Prof.
Ortega and Prof.
Kuo )
Our research in image and video compression is concerned with two major
areas, dealing with compression itself and with the system issues arising
in visual communications.
First we are interested in achieving higher performance than existing
algorithms, that is, reaching the same levels of perceptual quality at
higher compression ratios. This area of research includes projects dealing
with novel wavelet-based image coding techniques, some of which are being
considered within the JPEG 2000 standardization process. It also includes
research into post-processing and error concealment to improve the quality
of the decoded image or video data. Finally, we are considering methods
to improve the performance of existing compression standards, such as
MPEG or H.263, through the use of rate-distortion optimization techniques
and by designing algorithms that are scalable in terms of computation.
The second area of research deals with the system issues wich arise
in audio-visual communications. We are interested in adding functionalities
to compression algorithms so that they can support, for example, content
based retrieval, robust transmission over heterogeneous networks or burst
error wireless links. This work requires that trade-offs be made between
the compression performance and the necessary changes to the algorithms
to support the functionalities. Our research is being implemented in testbeds
for distributed audiovisual communication and multimedia information retrieval
over the Internet.
Research in the immersive audio lab is focused on signal processing
for loudspeaker-based 3-D audio reproduction, real-time video-based listener
tracking for adaptive crosstalk cancellation, real-time HRTF synthesis
for 5.1 channel rendering over two loudspeakers, and issues relating to
reproduction of high-quality audio for the desktop.
Research in medical imaging within SIPI focuses on aspects of medical
image acquisition, reconstruction, and processing. In particular, the
acquisition, reconstruction, and post-processing of MRI, PET, EEG, and
MEG. Our MRI group is currently developing fast and robust techniques
for high-field MRI (3 Tesla and higher). High field MRI provides high
SNR, but is limited by artifacts from RF inhomogeneity and
off-resonance, and by power deposition. We are developing new pulse
sequences that are robust to these new constraints, and image
reconstruction techniques that naturally compensate or correct image
artifacts. We also have funded projects related to the development of
Bayesian approaches to image reconstruction in positron emission
tomography (PET), spatio-temporal estimation methods for localizing
neural current sources from EEG and MEG data, and methods for the
automated segmentation and labelling of brain structures in 3D MR
images.
:
Our goal is to utilize the high data throughput potential of free space
optical signal propagation in the design and implementation of novel high
performance video and image processing systems, communication networks and
page-oriented memory interfacing. The performance of chips and subsystems
of current systems is limited by the relatively low data throughput of their
internal chip-to-chip and board-to-board interconnections. These limitations
will continue as the gap between on-chip and off-chip clock rates will grow
much wider in the forseeable future as CMOS performance increases. We study
free-space digital optics (FSDO) systems that use smart-pixel (SP) devices
to provide high-density, high-information-capacity, parallel optical interconnections
to electronic systems. Smart pixels are 'islands' of VLSI circuitry that
integrate electronic processing and free-space optical I/O channels. Part
of this research is the construction of working FSDO systems and the development
of enabling system components such as diffractive optical elements (DOEs).
The majority of the faculty of SIPI participate in the activities of
the Integrated Media Systems Center (IMSC). IMSC is a research center
concerned with the technological aspects of integrated media systems including
data compression, heterogeneous computing, distributed databases, wireless
communication, high-speed networks, augmented reality, and interactive
multimedia. The Center is a member of the National Science Foundation
Engineering Research Center Program and operates on the USC campus under
the auspices of the Viterbi School of Engineering.
Laboratories
The Signal and Image Processing Institute has a wide range of laboratory
facilities to support research and instruction in a variety of areas. Furthermore,
the SIPI Computer Center has a variety of computing and image processing
equipment used by the students, faculty and research staff.
| 