A foundation for logarithmic measures of fluctuating intensity in pattern recognition.

Independent analyses using Fisher information, optimal filtering, and information theory show that matched filtering images with hypothetical patterns in the logarithmic domain provides an optimal method for pattern recognition in the presence of signal-dependent noise arising from complex Gaussian fluctuations in the received fields. This provides a mathematical justification for the use of logarithmic units (i.e., decibels) in a variety of engineering applications.