This article presents a new MATLAB software, DigitSeis, that converts digital, raster images of analog seismograms to readily usable, digital time series using image‐processing techniques. It classifies important features of analog seismograms, such as time marks, to simplify the generation of continuous time series and uses some of this information to assign accurate timing. Geometrical distortions associated with various stages of seismogram creation, storage, and digitization are detected automatically and corrected. Although the software is written to minimize human input, DigitSeis provides interactive tools for challenging situations such as when traces cross. Finally, the effectiveness of the software is demonstrated with digitization of a long‐period seismogram from the Harvard‐Adam Dziewoński observatory recorded in November 1938. At least five significant teleseismic events are identified, and its spectral analysis shows no spurious features.