A software automation framework for image-typeface matching in graphic design

Abstract

This research proposes the framework for an automation tool that facilitates the graphic design process of image-font pairing or matching. Considering traditional graphic design principles, a multi-step software algorithm was developed to emulate the process of determining proportions and visual axes of both images and fonts. The algorithm then matches these visual markers using a decision hierarchy to produce a ranking of appropriate fonts from an existing font dataset. To test the algorithm, 8 benchmark images were selected with varying proportions and visual axes. To build the font data set, each image was manually analyzed through a traditional graphic design process and then two fonts per image with similar, matching characteristics were manually selected. The 8 benchmark images and 16 fonts were then used as inputs into the proposed matching software program. The results of the manually prescribed font-image pairings and calculated matches were then compared. Two images had the intended font in the top 4, two images had one of the intended fonts in the top 4, and 4 images had neither of the intended fonts in the top 4. An additional step in image-font pairing includes detail matching by determining curvature similarities. This detail analysis will affect the pairing outcomes and should be further investigated. This research began to analyze these details, and makes recommendations for continuing this work. Additional future directions for this work include incorporating a user-interface to the matching algorithm, introducing expert testing, and down-selecting the first font pool based on deviation.