Enhancements to a previously devised multiple model adaptive estimator for target image tracking are developed and analyzed: (1) allowing some of the elemental filters to have rectangular fields-of-view and to be tuned for target dynamics that are harsher in one direction than others, (2) considering both Gauss-Markov acceleration models and constant turn-rate models for target dynamics, (3) tuning dynamics noise strengths as a function of sensor-to-target range, and (4) devising an initial target acquisition algorithm to remove important biases in the estimated target template to be used in a correlator within the tracker. Particularly good adaptation responsiveness is demonstrated in the multiple model algorithm's ability to handle harsh maneuver onset, yielding performance essentially equivalent to that of the best artificially informed tracking algorithm.