Abstract: Oil, gas and water in a reservoir have different impedances, which may cause different reflection coefficients and waveforms. As a result, the value of correlation dimension is generally low. Seismic wave is the signal with definite bandwidth and the waveform is affected by all frequency components within the bandwidth, so the result could be unsatisfied if we calculate the correlation dimension directly. This paper proposes the method integrating wavelet packet with correlation dimension. Based on the feature that wavelet packet is suitable for decomposing high-low frequency component in detail, the high frequency component which can mirror details of seismic signals is extracted, and then the calculation of correlation dimension is made for the result given from frequency component processing. This method is applied to the middle oil layer of Triassic in Tahe area, and the calculated result is more satisfied.

Key words: wavelet packet transform, correlation dimension, reservoir rock, detection, high frequency, seismic data processing