Data Fusion BAUE Estimation of a deterministic vector, applications to image noise and blur reduction

Abstract

In this work we conceive centralized data fusion as a deterministic parameter estimation problem. Two different criterions are compared: best affine unbiased fusion rule (BAUE), and Maximum Likelihood for Gaussian measurement noise. Estimates are described in terms of their covariance matrices, the Cramer-Rao lower bound and simulations. The developed fusion rules are suited to two different image fusion cases: noise reduction under differently exposed images, and blur reduction based on lens response knowledge.