Estimating the inverse gradient

To optimize the fitting goal (74), module unwrap() uses the conjugate-direction method like the modules cgmeth() and invstack() . unwrapInverse 2-D gradient An open question is whether the required number of iterations is reasonable or whether we would need to uncover a preconditioner or more rapid solution method. I adjusted the frame size (by the amount of smoothing in Figure ) so that I would get the solution in about ten seconds with 400 iterations. Results are shown in Figure .

 

veshigh90
Figure 5 Estimated altitude.

To summarize, the input is the phase map Figure  and the output is the altitude map in Figure . Oddly, the input looks maybe nicer than the output because it already looks something like a contour plot. So if we have a beauty contest, then the input beats the output, but if you need to have the (normalized) altitude h(x,y), not the phase of e^ih(x,y), then you need to solve the least squares problem.