The time-domain induced polarization (IP) decay curve is a linear superposition of IP signal from individual pores. It can be transformed to relaxation time spectrum related to the pore structure and can be used to estimate the permeability of the argillaceous sand. So the IP logging is a useful tool for reservoir characterizing. The singular value decomposition (SVD) method can be used to extract the relaxation time spectra from the decay data with high signal-to-noise ratio (SNR). This method is unstable and invalid for the practical downhole use because of the noisy decay data. In this study, a method of regularization was applied to extracting continuous IP relaxation time spectra from simulated and real decay data. Influence of the regularization factor, SNR and relaxation time arrangement on the spectra was also discussed. Firstly, for the simulated decay data with noise, the spectra become smoother with increasing regularization factor. There is an optimal value which can be used to get the most reasonable spectrum. The optimal factor decreases linearly with increasing SNR on a log-log scale. This relationship can be used to predict the optimal factor from a known SNR. Secondly, the influence of relaxation time arrangement on the spectra shows that the suitable relaxation distribution point ranges from 32 to 64. Finally, the developed algorithm and the prediction of the optimal factor are very suitable for the inversion of the spectra of the real data.