The WAVES Lab is developing new approaches for compressed sensing and rank minimization
for sparse signals, in general, and for digital imaging in particular. This research,
partly funded by NSF, is enabling novel algorithms for demosaicking, deblurring,
denoising, super-resolution, and visual coding. It also includes a major collaboration
with Kodak Research Labs. The utility of hybrid (sparse and dense) projections that
lead to low-complexity decoding algorithms are being explored. This framework also
leads to channel-coding like compressive sampling. Apllications include video, speech,
networking, and pattern recognition.
Social network analysis
In collaboration with other research groups, the WAVES Lab is pursuing an intriguing
technology direction for the analysis of on-line social networks and their underlying
infrastructure and services. We are developing new graph transforms, signal processing,
information-theoretic, and machine learning tools for the analysis and understanding
of massiv social network graphs and services. A core component of this NSF funded
multidisciplinary research is close interaction with social scientists and other
experts in related fields.
Visual retargeting & summarization
New approaches for content-aware visual "retargeting" are being developed. A variety
of display devices that prohibit maintaining the original aspect ratio of the captured
visual content are considered. Multiresolution approaches that are robust to a
variety of visual content distortion are being designed and analyzed. Extensions
of these approaches to 3D visuals and displays are being investigated. Novel video
summarization frameworks are being explored.
Reliable & stable wireless multimedia
Reliable, efficient and stable wireless link-layer protocols are being designed,
analyzed and implemented for high-end emerging multimedia applications over wireless
networks. These protocols, designed by support in part from NSF and industry, provide
joint reliability and stability for both delay constrained real-time video applications
and traditional applications that are built on TCP/IP. Target applications include
multimedia streaming, telemedicine/health monitoring, surveillance, and gaming.
The WAVES Lab ACE Protocol
network coding & network channel coding
Migrating coding functions from end nodes into intermediate nodes within the network
represent a major paradigm shift with improvements in throughput, delay, and/or
reliability. Based on support from two NSF grants, the WAVES Lab has developed advanced
network-embedded source and channel coding approaches over network topologies that
are representatives of next generation Internet and wireless networks, sensor networks,
peer-to-peer networks, and ad-hoc networks.
Optimal Distribution of Source and Channel Coding
Functions over Random Networks using Network
Coding and Network Channel Coding
practical distributed video coding
Practical Distributed Video Coding (PDVC) includes designing and analyzing
new practical approaches for the mapping of Distributed Video Coding (DVC) algorithms
into low-complexity low-power visual sensors and devices. Applications of PDVC onto
emerging multi-view/3D video are being explored. The WAVES Lab also developed novel
approaches based on multi-hypothesis distributed visual coding. Partly supported
by NSF, this effort includes the redesign of state-of-the-art coding schemes, such
as Low-Density-Parity-Check (LDPC) and Polar codes for DVC.
