Comparison of nonlinear joint transform correlator and nonlinearly transformed matched filter based correlator for noisy input scenes

We compare the performance of the nonlinear joint transform correlator, the nonlinearly transformed filter based correlator, and the conventional linear correlator. The pattern recognition systems are compared in the areas of correlation peak, peak to sidelobe ratio, and correlation width. We show that for some noisy input scenes, the linear correlator produces an autocorrelation peak that is smaller than the maximum cross-correlation signal, resulting in false signal detection. On the other hand, compared to the linear correlator, both types of nonlinear correlators produce better correlation performance.