Abstract

A joint transform power spectrum (JTPS) usually has a very high dc component; an attempt to record the whole intensity spectrum produces a low-pass effect that results in a wider correlation peak. There are various thresholding schemes to improve the performance, such as binarization of the JTPS, or other methods encountered in practical situations, such as the hard clipping effect from the limited dynamic range of the square law detector or simply the dc blocking scheme that blocks the dc part of the JTPS. To find an optimum thresholding scheme, we have compared the performances of all these thresholding methods, using various inputs imbedded in an additive Gaussian noise environment. Our computer simulation results show that the dc blocking method is comparable in performance with that of the binarization method and is much simpler from the experimental point of view.

© 1991 Optical Society of America