Causal and semicausal stochastic elds may be rewritten in a state-space representation. Local changes in these elds may be regarded as time-varying changes in the parameters of this model. Using a Kalman lter bank, the generalized likelihood ratio may be computed if these changes are known to be in a certain set. If only some scaling parameters are unknown, exact expressions for the generalized likelihood ratio may be found.
|Publisher||University of Twente, Department of Applied Mathematics|