| A generalized pattern recognition system utilizing normalized cross-correlation (NCC) via fast Fourier transforms (FFT) is investigated. This system is applied to the human face recognition problem, in an attempt to define success ratios, spatial power bandwidths, and quantization requirements. Statistical properties of the facial problem are investigated.; Charge-coupled device (CCD) architecture is applied to the NCC system. A set of CCD simulation subroutines is compiled to provide a programmable model of CCD architectures along with the physical properties of the devices themselves. Physical effects within CCDs (charge decay, transfer inefficiency, quantization, dark current, etc.) are judged as to their effect on the NCC system performance.; Finally, optical preprocessing for NCC systems is discussed, along with a general discussion of optical CCD structures for other transforms. |
