Wave separation is a main issue in seismic signal processing. Two waves having different polarizations can be separated if the polarization of each wave is known. This is performed by an oblique polarization filter. This technique developped recently has a crucial issue. It requires the estimation of the polarization of each wave, which can be a difficult task. This paper presents the use of the time-scale plane to estimate the polarization of the waves. The module of the continuous wavelet transform allows to detect the different waves, and the phase wrapping matrix enables to detect the interferences between waves. The algorithm is illustrated on synthetic data and successfully tested on real seismic data. It succeeds in separating two waves overlapping on the time-scale plane.