Bispectral analysis of two-dimensional random processes

The extension of bispectral analysis from one-dimensional random processes (e.g. time series) to two-dimensional random processes is presented. Use of the symmetry properties of the bispectrum reduces the number of computations considerably. Numerical simulations demonstrate the ability of two-dimensional bispectral estimation to detect quadratic phase coupling between waves traveling in different directions. By windowing the data, bicoherence leakage can be reduced.>