Introduction to Second Kind Statistics: Application of Log-Moments and Log-Cumulants to the Analysis of Radar Image Distributions

Statistical methods classically used to analyse a probability density function (pdf) are founded on the Fourier transform, on which useful tools such the first and second characteristic function are based, yielding the definitions of moments and cumulants. Yet this transformation is badly adapted to the analysis of probability density functions defined on R+, for which the analytic expressions of the characteristic functions may become hard, or even impossible to formulate. In this article we propose to substitute the Fourier transform with the Mellin transform. It is then possible, inspired by the precedent definitions, to introduce second kind statistics: second kind characteristic functions, second kind moments (or log-moments), and second kind cumulants (or log-cumulants). Applied to traditional distributions like the gamma distribution or the Nakagami distribution, this approach gives results that are easier to apply than the classical approach. Moreover, for more complicated distributions, like the K distributions or the positive αstable distributions, the second kind statistics give expressions that are truly simple and easy to exploit. The new approach leads to innovative methods for estimating the parameters of distributions defined on R+. It is possible to compare the estimators obtained with estimators based on maximum likelihood theory and the method of moments. One can thus show that the new methods have variances that are considerably lower than those mentioned, and slightly higher than the Cramér-Rao bound.

[1]  M. Nakagami.  Statistical Methods in Radio Wave Propaga-tion , 1960 .

[2]  H. Cartan,et al.  Théorie élémentaire des fonctions analytiques d'une ou plusieurs variables complexes , 1961 .

[3]  Sur certaines propriétés de la distribution gamma généralisée , 1967 .

[4]  Z. A. Lomnicki.  On the Distribution of Products of Random Variables , 1967 .

[5]  M. Kendall,et al.  Kendall's advanced theory of statistics , 1995 .

[6]  C. Oliver.  Optimum texture estimators for SAR clutter , 1993 .

[7]  C. L. Nikias,et al.  Signal processing with fractional lower order moments: stable processes and their applications , 1993, Proc. IEEE.

[8]  R. D. Pierce.  Application of the positive alpha-stable distribution , 1997, Proceedings of the IEEE Signal Processing Workshop on Higher-Order Statistics.

[9]  Jean-Marie Nicolas,et al.  Images de radar à Ouverture synthétique : transformée de Mellin et multirésolution , 1997 .

[10]  Speckle well modeled by Mellin transform , 1998 .

[11]  Les moments d'ordre inférieur: Principes et application au filtrage des images RSO , 2000 .

[12]  J. Zerubia,et al.  Modelling SAR images with a generalisation of the Rayleigh distribution , 2000, Conference Record of the Thirty-Fourth Asilomar Conference on Signals, Systems and Computers (Cat. No.00CH37154).

[13]  Aldo Tagliani,et al.  Recovering a probability density function from its Mellin transform , 2001, Appl. Math. Comput..

[14]  Aldo Tagliani,et al.  Numerical inversion of the Mellin transform on the real line for heavy-tailed probability density functions , 2002, Appl. Math. Comput..