C. Hu, R.D. Hersch
Proc. Int'l Conf. on Image Processing (ICIP'99), Kobe, Japan, Vol. 4, 1999, 348-351
In today's digital computers, phototypesetters and printers, typographic fonts are mainly given by their outline descriptions. Outline descriptions alone do not provide any information about character parts like stems, serifs, shoulders, and bowls. But, in order to produce the best looking characters at a given size on a specific printer, non-linear operations must be applied to parts of the character shape. At low-resolution, grid-fitting of character outlines is required for generating nice and regular raster characters. For this reason, grid-fitting rules called hints are added to the character description. Grid-fitting rules require as parameters certain characteristic points within the shape outlines. In order to be able to detect these characteristic points in any given input font, a topological model representing the essence of the shapes found in topographic latin typefaces is proposed. This model includes sufficient information for matching existing non-fancy outline fonts to the model description. For automatic hint generation, applicable hints are joined to the topological model description. After matching a given input shape to the model, hints which can be applied to the shape of the given font are taken from the model and added to the matched character outline description. Furthemore, a structural description of individual letter shape parts using characteristic model points can be added to the model. Such a description provides knowledge about typographic structure elements like stems and bowls.
Download the full paper: PDF 232 KB