Most research in computer vision is largely related to theoretical studies and algorithm development, but is generally isolated from hardware and computer architecture, which creates a barrier for deploying real world products for the mass market. Thus, it is desirable to have vision algorithms that are suitable for hardware implementation.

We have been conducted in-depth investigation on a set of vision problems from both software and hardware perspectives. One is the possibilities of hardware implementation for face detection. With Dr. J. Crenshaw and his colleagues at Motorola Labs, we have designed a hardware pipeline design that is verified on a Cyclone II FPGA.

In addition, collaborating with Profs. A. K. Katsaggelos, A. Choudhary, G. Memik and S. Memik at Northwestern, we have been working on high-performance designs for change detection. Our goal is to develop novel algorithms and hardware schemes for real-time modeling and change detection on large-scale streaming data sets.

[1] Ming Yang, James Crenshaw, Bruce Augustine, Russell Mareachen, and Ying Wu, "Face Detection for Automatic Exposure Control in Handheld Camera", in Proc. IEEE Int'l Conference on Computer Vision Systems (ICVS'06), New York City, Jan. 2006.  [PDF]
 

Return to Research