Len Bruton's Research Home Page
Multidimensional and Multimedia Signal Processing
logo by Don Dansereau
INDEX
The Multidimensional Signal Processing (MDSP) Group Home Page
- synthesis, design and real-time VLSI circuit implementations of 2D,3D and 4D filters, including 3D and 4D filter banks
- enhancement of video sequences using 2D, 3D and 4D filters
- real-time enhancement and jamming of sampled broadband spatio-temporal plane waves using analog or digital 2D and 3D filters
- adaptive 2D and 3D filters
- 2D and 3D beamforming using filters
- 3D modeling of faces for videoconferencing
- real-time enhancement of damaged films
- object tracking and identification in video sequences
- data hiding in and watermarking of video sequences
- object extraction in 4D lightfields using 4D filters
For broadband sampled-plane wave filtering, we have recently shown that high performance IIR 3D cone filter transfer functions may be implemented using computationally-efficient 3D filter banks and so we are currently exploring ways of implementing these filter banks using Xilinx Field Programmable Gate Arrays (FPGAs) and Texas Instruments 6000 Series Processors.
For high performance narrow FIR 2D fan and 3D cone filters, we have recently demonstrated that the computational overhead associated with conventional 2D and 3D input-output convolutions may be very significantly reduced by employing 'a priori' knowledge of the orientation of the 2D/3D transients of the unit impulse response. FPGA single chip implementations have been shown to be feasible at temporal sampling frequencies of over 100 MHz. We are exploring different types of FPGA single-chip 2D, 3D and 4D filter architectures.
Our recent current work on adaptive 3D IIR filters has demonstrated that we can adaptively detect and jam a broadband spatio-temporal plane wave as its direction of arrival (DOA) varies slowly with time relative to the receiving array.
We have implemented the first reported IIR 4D 'cone' filter and shown that it may be used to extract objects within 4D lightfields on the basis of their distance from the viewing plane. This work has applications in computer vision.
An acoustic analog 2D spatio-temoral RC-Active Beam Filter circuit has been constructed in prototype form for the purpose of selectively filtering audio signals based on their direction of arrival and using microphone arrays . The method has the potential for very low cost.
A software package has been developed for watermarking or hiding other types of data in video using 3D filters. The demonstration allows watermarks to be embedded in 240x320 avi-formatted video in close to real-time.
Download Current and Past Publications
Most of my publications are available for download in PDF format from my Publications List and are organized in reverse chronological order.
Applications of Multidimensional (MD) Filters
MD filters operate on MD images. Typically, the independent variables of 2D images are (space-space) or (space-time). The independent variables of 3D signals are typically (space-space-space) or (space-space-time).
Typical practical problems are:
the enhancement of 2D visual images
the selective filtering of plane waves in 2D and 3D images using fan or cone filters
the extraction of objects from 4D lightfields using 4D filters with applications in computer vision.
Although methods have been available for several decades for synthesizing both FIR and IIR 2D and 3D filter transfer functions, until recently the advent of low-cost hardware inhibited the implementation of specific real-time filter architectures in a VLSI circuit environment. It is now possible to employ modern FPGA devices and other VLSI circuits to implement real-time 2D, 3D and 4D filters having extraordinary frequency selectivity at temporal sample frequencies up to many hundreds of megahertz, suggesting a myriad of real-time engineering applications from seismic image enhancement to the radio-frequency enhancement and jamming of plane waves. Such potential applications include directional audio systems, ultrasonic imaging in remote sensing and biomedicine, the extraction of objects in computer vision applications, such as robotics, and wireless beamforming.
Many international, national, regional and university awards and honors have been received for our research. A complete list of awards and honors is available here. They include the national 1991 Manning Principal Award for Innovation, election to the Royal Society of Canada and election as a Fellow of the IEEE..
A video of interviews with past winners of the Manning Principal Award is available at Past Winners
My research is currently supported by Natural Sciences and Engineering Research Council (NSERC) of Canada and the University of Calgary.
Further details are available from our the Multimensional
Signal Processing Group Home Page.