When the oil spills on water are passively detected by differential polarization Fourier transform infrared (FTIR) spectroscopy in the band of 3~5 μm, the measured differential polarization spectrum will be a mixed spectrum of the strong atmospheric absorption signal and the minute signal of oil pollutant. It brings complexity to the identification of oil spills on water. In addition, the thickness distribution and the surface coarseness of oil film on water change with the influence of environmental factors and the tension of oil film. This makes the content of effective spectral information, which is contained in the differential polarization spectrum of oil spills on water, change continuously during the process of measurement. Using this characteristic, an algorithm based on a fast fixed-point algorithm for principal component analysis (FastPCA) is proposed for preprocessing the differential polarization spectrum of oil pollutant on water. Experimental results show that this algorithm can separate the spectral information of oil pollutant from the differential polarization spectrum with strong atmospheric absorption signal. The reconstructed spectral characteristic signal of oil pollutant can be used for the further qualitative and quantitative analyses.