One of the difficulties in seismic phase unwrapping occurs at the frequency
components with energy close to zero, where the final. unwrapped phase fun
ction should have +/-pi discontinuities. At those frequency components, pri
ncipal values of the wrapped phase spectrum and in turn the values of the f
inal unwrapped phase function are contaminated severely by numerical errors
. These errors also propagate along the frequency axis as the unwrapped pha
se function is calculated recursively. This paper presents a robust phase u
nwrapping method. Unwrapped phase function is defined as the sum of phase i
ncrements, where the phase increment of adjacent frequency components is ev
aluated in terms of the difference of principal values of the wrapped phase
spectrum. In the vicinity of +/-pi discontinuities, appropriate treatment
of phase increments is required, preventing the numerical errors from propa
gating along the frequency axis.