Stability is the key to inverse Q-filtering. In this paper we present a stable approach to inverse Q-filtering, based on the theory of wavefield downward continuation. It is implemented in a layered manner, assuming a layered-earth Q model. For each individual constant Q layer, the seismic wavefield recorded at the surface is first extrapolated down to the top of the current layer and a constant Q inverse filter is then applied to the current layer. When extrapolating within the overburden, a stable wavefield continuation algorithm in combination with a stabilization factor is applied. This avoids accumulating inverse Q-filter errors within the overburden. Within the current constant Q layer, we use Gabor spectral analysis on the signals to pick time-variant gain-constrained frequencies and then deduce the corresponding gain-constrained amplitudes to stabilize the inverse Q-filtering algorithm. The algorithm is tested and verified application to field data.
