A robust method is suggested for the classification of the porous samples of a given material in terms of permeability heterogeneities. Datasets from rate-controlled immiscible displacement experiments in heterogeneous soil columns and dynamic simulations of oil/water drainage in permeability networks are employed to examine the effects of heterogeneities on the transient response of the pressure drop across a porous medium. The multi-level wavelet analysis of the pressure drop signal up to the best level enables us to correlate the variability and spatial correlations of permeability field with the statistics of the wavelet energy and frequency, which could be used as parameters for the pattern recognition of heterogeneity at the scale of measurement. (C) 2010 Elsevier Ltd. All rights reserved.