A straightforward algorithm for estimating p at (t,x) is to scan the objective function En(t,x,p) over a reasonable range of p, and then find the minimizer of the objective function with a 1-D search. The dip obtained with this algorithm is an unconstrained solution to the problem because p can be any arbitrary function of (t,x). This unconstrained solution is sensitive to noise in the data; it may be completely wrong when the data are insufficiently sampled.