Removing ths statistical bias from three-dimensional noise measurements

摘要

The three dimensional noise model (3D noise) is a widely used model for characterizing noise in thermal imaging system. In this model, a sequence of images of a uniform background are acquired, and organized in a three dimensional matrix. This matrix is then decomposed into eight orthogonal noise components that can be assessed individually to yield an understanding about the magnitude and source of noise in a given system. In a previous paper we showed that the operators used to estimate the magnitude of the 3D noise in a system are biased statistical estimators that lead to systematic errors when measuring system noise. Here we provide new definitions for the noise estimators that enable removal of the statistical bias, and accurate estimation of system noise using the 3D noise model.