Research
Interests |
|
Real time video/image processing Although available computational power increases every day, high computational cost of image processing algorithms for some applications remains too high for real time processing. Medical imaging is an special case because of cost, precision and complexity involved, for instance, in multiple modalities image registration, pattern matching, image segmentation, etc. The development of techniques to improve the computational cost and reduce the time needed to obtain useful information for diagnostic remains an interesting challenge. Some links: IVPLC Lab
|
|
Programmable logic Issues related to real time signal processing are very well addressed by programmable devices. Advantages such as development time, cost, speed, integration density, flexibility capacity for parallel implementations, etc. set programmable devices as the ideal choice over general purpose DSP processors and VLSI implementations. Abundant literature reports the use of programmable devices in image processing systems. Following this path our research is oriented to efficient implementation of high costly image processing algorithms using Field Programmable Gate Arrays (FPGAs). Some links:Xilinx, Celoxica, Nallatech, Andraka, EECE 238 course
|
|
Numerical Optimization Numerical optimization algorithms are present in every digital signal processing application. Particularly important are the methods to find sparse representations of a signal in overcomplete dictionaries. Such a search leads to a large scale optimization problem. The analysis and processing of a signal is simplified when a sparse representation is available, thus the importance of such a search. When implementing in hardware such methods, the designer has to deal with two main issues: the iterative nature of the algorithms and its sensitivity to quantization errors. Both issues make the implementation of such algorithms an interesting challenge. Some links: Neos Guide
|