Compression of digital elevation maps using nonlinear wavelets
We developed a compressed representation for digital elevation maps using min– and max-lifted wavelets for that supported fast search and retrieval. Based on this representation, we built a GUI-based prototype system.
This project was supported by the Office of Naval Research and ran from 2000 to 2003. Numerous conference papers and one journal paper were published documenting this work.
Intelligent compression for remote sensing
The main objective of this project was extraction of useful information directly from the compressed representation of the signal (in this case, images & video). We considered here the problems of spatial object detection, localization, and classification in JPEG and wavelet-compressed imagery as well as extended-object target tracking using H.264 motion vector information. A number of conference papers have been published documenting our research results.
This project was sponsored by Sandia National Labs and was active from 2001-2003.
Passive polarimeteric imagery classification study
In this project, we have studied the utility of passively gathered polarimetric imagery for spatial object detection and classification. We have developed a multi-look strategy using a physics-based reflection model which allows us to extract geometrically invariant feature vectors for classification, segmentation, and detection.
This work is supported by the National Geospatial Intelligence Agency (NGA). The initial project ran from 2004-2006 and we are currently working on a follow-on project which started in 2006. To date, this project has produced numerous conference papers and two journal papers.
Efficient audio compression with perceptually embedded scalability
The goal of this project was to develop scalable audio compression algorithms that were capable of performing well over a range of bitrates. In particular, we designed the coder to have fine-grained scalability starting from a rate of 8 kb/s where each increasing bit allocation was made so that a perceptual criterion was optimized. As part of this research, we also studied in detail the human perception of audio quality in the high to moderate impairment range.
This project was funded by a National Science Foundation Early Career Grant from 2002 to 2008 and has resulted in four journal publications and numerous conference publications. Matlab-based software that has been developed as part of this project for both audio coding and perceptual quality analysis can be downloaded here.
Distributed source coding using bitstream-based detection and classification
In this project, we are developing practical distributed compression algorithms for wireless sensor networks that extract inter-sensor correlations on-the-fly and exploit them to achieve more efficient bandwidth utilization. Our specific focus is on video and audio data, and correlation extraction is performed using the compressed bitstream so as to reduce encoding complexity as much as possible.
This work is sponsored by the Army Research Office and has thus far led to 2 published conference papers.