Detecting upper outliers in small gamma samples: A comparison of techniques

LUCINI, MARÍA MAGDALENA; FRERY, ALEJANDRO CESAR; MEDEIROS, ANTONIO; ROSSO, OSVALDO

Conferencia; Asia Pacific Conference on Synthetic Aperture Radar; 2015

IEEE

A common assumption for Synthetic ApertureRadar (SAR) data, is that the intensity return from texturelessareas follows a Gamma law with mean > 0 and L > 0looks. Many image processing techniques need to estimate theseparameters using small samples. Unfortunately, the presence ofdiscrepant observations in SAR data occurs frequently, evenwhen dealing with small samples. This is mostly caused bythe presence of a small strong backscatterer as is the case of,for instance, a corner reflector. Processing techniques based onthe estimation of the parameters of the sample distribution arehighly influenced by the presence of such outlying observations.Therefore, it is of paramount importance to identify them beforeapplying techniques based on parameter estimation. We discusstest statistics designed to detect the presence of an outlyingobservation in a gamma sample, and we propose a new test basedon an empirical estimate of the underlying distribution.