Previous studies have verified the feasibility of using ambient noise in water pipeline networks as natural signals to assess pipeline condition passively. This passive technology uses correlation peaks of the ambient noise as characteristics to identify pipe faults and deterioration, and it requires the noise power spectrum density (PSD) to be adequately whitened. However, conventional noise whitening methods cannot provide sufficient whitening for the ambient noise in water pipelines due to the significant noise PSD variations that are unevenly distributed across the wide relative bandwidth of the noise signals. This paper proposes a tailored whitening process for water pipeline ambient noise by applying logarithmic transforms on both the magnitude and frequency axes to compresses the large dynamic range of the PSD magnitude and bandwidth. Numerical simulations and field experiment results have verified that the proposed method can effectively whiten the noise signals in water pipelines, and accurately restore the amplitude of correlation peaks to be consistent with the theoretical derivation using white noise assumptions. The theoretical derivation also reveals the relation between the amplitudes of correlation peaks and the noise spatial distribution, fault distance and its reflection/transmission coefficients. The feasibility of estimating the reflection coefficient of pipe faults from the correlation results with appropriately whitened noise is also investigated in this paper.