The goal of letter fitting is to make the letters in a word appear equidistant from one another, while ensuring that adjacent letters do not collide. In principle, this formulation should be sufficient to allow letter fitting to be fully automated. That this has not happened is largely due to the difficulty of measuring the distance between two letters in a way that accords with human perception. Here I propose a simple approach to measuring the optical distance between two letters, grounded in a theoretical understanding of human vision. To measure the distance between two letters, we first apply optical blur to the letterforms, to arrive at their perceived shapes. Then we identify and fill in counters using the watershed algorithm. Then we measure the area between the two resulting shapes, ignoring regions where one letter extends above or below the other. Finally, we divide the area by the length of the “spine” running through the middle of it. I demonstrate that a letter fitting algorithm incorporating this definition of optical distance produces visually appealing results at a range of optical sizes, for multiple writing systems including the Latin alphabet and Japanese kana.
An Optical Approach to Automated Letter Fitting
