Towards a universal auto-hinting system for typographic shapes

The article presents a simple method for the automatic recognition and hinting of character structure elements such as horizontal and vertical stems. Stem recognition is based on successive steps such as extraction of straight or nearly straight contour segments, detection of hidden segments, merging of original and hidden segments into larger segments, sorting of segments into classes according to their slopes and finally, composition of black and white stems. Reference values required for character hinting purposes are obtained by evaluating the regularity of the font through statistical analysis of features such as stem widths and stem angles. Knowledge about the location of stems and analysis of outline parts between stems is used in order to produce automatically appropriate grid constraint rules (hints). The presented outline analysis and stem extraction techniques are very general and may be applied to non Latin characters as well