LOCALLY NONLINEAR MATCHED FILTERING

A new matched filter for pattern recognition is introduced. Previous r esearchers have introduced matched filters that are nonlinear function s of the spectrum, for which the classical matched filter is divided b y some power m of the spectrum of interest. We further generalize this filter by making the power m a function of the spatial frequency; thi s permits the design of filters that combine the advantageous properti es of matched filters, phase-only filters, and inverse filters without the corresponding disadvantages. The first computer experiments indic ate that the new filter yields sharper correlation peaks and better di scrimination than the matched filter and the phase-only filter and yie lds more robustness against noise.