A simple method for digitizing of infrared (IR) spectrum from scanned images of published data is proposed and tested for amorphous and polycrystalline silicon alloys. The method uses the following special constraints of IR spectral data: (a) scanned image should have only two gray levels (black and white); (b) each point on the graph should be single valued; (c) frequency distribution of non-background pixels of columns should give a maximum, at the number of scan lines in the images. These criteria along with the closest distance approach are used in programs for identifying IR spectra and found to be successful in more than 95% of the cases. After identification of the lines, base line correction, normalization, and noise reduction techniques were applied before recording the data to the magnetic media. This method can be used for purposes other than forming an IR reference library, like archiving, regraphing, or comparing any spectral data.