We show that we can use the linearized WEMVA operator to invert all three types of image perturbations. We find that Rytov WEMVA is possible if we unwrap phases, but fails without.
The differential image perturbation can be used for arbitrarily large slowness anomalies, but its accuracy decreases with deviation from the background image. For simple cases, the differential image perturbation method is equivalent with a phase-unwrapped Rytov method.