A method of subtraction and normalization of IR spectra (MSN-IR) was developed and successfully applied to extract mathematically the pure paint spectrum from the spectrum of paint coat on different bases, both acquired by the Attenuated Total Reflectance Fourier Transform Infrared (ATR-FTIR) technique. The method consists of several stages encompassing several normalization and subtraction processes. The similarity of the spectrum obtained with the reference spectrum was estimated by means of the normalized Manhattan distance. The utility and performance of the method proposed were tested by examination of five different paints sprayed on plastic (polyester) foil and on fabric materials (cotton). It was found that the numerical algorithm applied is able – in contrast to other mathematical approaches conventionally used for the same aim – to reconstruct a pure paint IR spectrum effectively without a loss of chemical information provided. The approach allows the physical separation of a paint from a base to be avoided, hence a time and work-load of analysis to be considerably reduced. The results obtained prove that the method can be considered as a useful tool which can be applied to forensic purposes.