Bayesian theory for target location in noise with unknown spectral density

A Bayesian approach adapted to practical detection and location tasks of a target in additive Gaussian noise with unknown spectral density is developed and studied. The relevance of this theory is first discussed in comparison with a maximum a posteriori solution that has been recently developed [J. Opt. Soc. Am. A15, 61 (1998)]. The analysis is first performed without considering a particular prior for the spectral density of the noise. General results of the Bayesian approach are thus provided as well as properties of its first-order development, which corresponds to the so-called nonlinear joint-transform correlation frequently used in optical correlators. It is demonstrated that the kernel of the nonlinear filtering is an increasing function of the sum of the spectral density of the reference object and of the input image. Furthermore, it is shown that a power-law mathematical form of the nonlinear filtering is directly related to assumptions on the asymptotic behavior of the prior density probabilities of the unknown spectral density of the noise. These properties constitute new theoretical results in the context of statistical theory concerning the use of nonlinearities in optical correlators.

[1]  B Javidi,et al.  Basic properties of nonlinear global filtering techniques and optimal discriminant solutions. , 1995, Applied optics.

[2]  Bahram Javidi,et al.  Minimum mean-square-error filter for pattern recognition with spatially disjoint signal and scene noise , 1993 .

[3]  B Javidi,et al.  Nonlinear joint-transform correlation: an optimal solution for adaptive image discrimination and input noise robustness. , 1994, Optics letters.

[4]  B Javidi,et al.  Optimum receiver design for pattern recognition with nonoverlapping target and scene noise. , 1993, Optics letters.

[5]  Philippe Réfrégier,et al.  DECISION THEORY APPROACH TO NONLINEAR JOINT-TRANSFORM CORRELATION , 1998 .

[6]  P Réfrégier,et al.  Compact photorefractive correlator for robotic applications. , 1992, Applied optics.

[7]  Philippe Réfrégier,et al.  Comparison of the performance of linear and nonlinear filters in the presence of nonergodic noise , 1997 .

[8]  F Goudail,et al.  Influence of nonoverlapping noise on regularized linear filters for pattern recognition. , 1995, Optics letters.

[9]  J. Goodman,et al.  A technique for optically convolving two functions. , 1966, Applied optics.

[10]  Vincent Laude,et al.  Bayesian target location in images , 1997 .

[11]  B. Javidi.  Nonlinear joint power spectrum based optical correlation. , 1989, Applied optics.

[12]  Ph. Réfrégier.  Application of the stabilizing functional approach to pattern-recognition filters , 1994 .

[13]  A. B. Vander Lugt,et al.  Signal detection by complex spatial filtering , 1964, IEEE Trans. Inf. Theory.

[14]  B Javidi,et al.  Limitation of the classic definition of the correlation signal-to-noise ratio in optical pattern recognition with disjoint signal and scene noise. , 1992, Applied optics.

[15]  S Vallmitjana,et al.  Nonlinear filtering in object and Fourier space in a joint transform optical correlator: comparison and experimental realization. , 1995, Applied optics.

[16]  M J Yzuel,et al.  Nonlinearity effects in the pure phase correlation method in multiobject scenes. , 1994, Applied optics.

[17]  Vitaly Kober,et al.  Accuracy of location measurement of a noisy target in a nonoverlapping background , 1996 .

[18]  L. Pichon,et al.  Dynamic joint-fourier-transform correlator by Bragg diffraction in photorefractive Bi12SiO20 crystals , 1981 .

[19]  Bahram Javidi,et al.  Design of filters to detect a noisy target in nonoverlapping background noise , 1994 .

[20]  P. Réfrégier.  Filter design for optical pattern recognition: multicriteria optimization approach. , 1990, Optics letters.