MAP estimation using a Viterbi approach with continuous-state representation

In most target tracking references, the problem of detecting and tracking small dim objects in image sequences is cast as a nonlinear maximum a posteriori (MAP) state estimation problem on a dynamical system. In these applications, it is a state sequence that defines the trajectory to be estimated. Although discrete-state Viterbi algorithm approaches have been used when the underlying process is Markov, their implementation is limited to very small arrays due to high computational requirements. The authors pursue an alternate approach by approximating the terms in the recursive metric calculation with piecewise linear Lagrange interpolation. This approach results in a continuous-state Viterbi algorithm which is computationally tractable for nonlinear MAP estimation