Feature extraction algorithms using FPGA technology

This paper reviews the implementation of two feature extraction algorithms built using FPGA technology. The algorithms, based on moment invariants and the rapid transform, have been chosen to illustrate the potential of this technology for implementing simple but computationally intensive algorithms for object identification, scene matching and character recognition. In the past such circuits have needed either custom VLSI chips, or dedicated software on DSPs or workstations. The circuits described manipulate either binary or 256 grey scale images from a 128/spl times/128 pixel digital camera in real time (30 frames per second) and represent front end processors for flexible high performance image processing systems.