During measuring the thickness of metallic foil with tandem differential white light interferometry, a new data processing method for white light interference spectrum based on the Minimum Mean-square Error (MMSE) was proposed to overcome the poor anti-noise capability of traditional extremum method. In the new method, the spectrum measured value was firstly normalized to eliminate the intensity modulation, and an initial value of Optical Path Difference (OPD) was calculated by extremum method. The mean square deviation between the normalized measured spectrum and the theoretical spectrum from the initial value was calculated. When the estimated value of OPD is closest to the actual one, the mean square deviation will be minimum. The optimal estimated value was obtained by means of iteration. Emulation and experiments show that the data processing method has higher precision and better Signal-to-noise Ratio (SNR) than the traditional extremum method.