Abstract
Selecting an appropriate typeface is crucial in typography, web design, and other applications where text readability is essential. A key concept of this topic is legibility, the quality that shows how easy it is to recognize the letters of a particular font set. Previous works have measured legibility by human experiments, which has several limitations; for example, the methodology and circumstances were not entirely uniform, and the results may be distorted by the fatigue of the test subjects. This paper presents a new method using self-developed software to substitute human measurements in legibility testing. The software simulates the human retina's distortion effects (direction-dependent acuity) on test images of the assessed font's characters. Then, its output is analyzed using optical character recognition software. By integrating these techniques, we model the optical, biological, and cognitive steps of human character recognition as well. Although the simulation is imperfect, the software can perform significantly more measurements than human experiments with higher uniformity and give reproducible legibility information about significantly more fonts in various circumstances. In addition to the two scaling methods used in the literature (x-height, font-height), the tests are also performed with two other self-developed scaling methods, which provide a fairer comparison in the case of non-standard character types. This paper contains the legibility measurement results for 22 fonts under various simulated scenarios. The derived font ranking aligns closely with findings from prior human-based studies, demonstrating the robustness and reliability of the proposed method. Moreover, this approach provides valuable insights into font legibility across a broader spectrum of use cases, highlighting its potential for practical applications in typography and design.