As the demand for media content increases, interest in the font of text that makes up digital media content is also increasing to effectively deliver the content. Because fonts are used in various fields, fonts with various characteristics are created...
As the demand for media content increases, interest in the font of text that makes up digital media content is also increasing to effectively deliver the content. Because fonts are used in various fields, fonts with various characteristics are created, and some people want to make their own handwriting into fonts. Handwritten fonts by humans show the person's characteristics well, but they are often not readable due to problems such as crushed letters or inconsistent sizes.
Existing image-based font generation models have the disadvantage of being difficult to adjust detailed font styles such as letter size, stroke shape, and structure because they try to generate fonts in a certain style through small input data. Therefore, this paper proposes a model that can generate fonts based on noise to enable font-style microcorrection.
FontFusionGAN (FFGAN), a model to be introduced in this paper, creates a new font that mixes handwriting with font styles of the Myeongjo and Gothic series when handwriting is made into fonts. FFGAN can produce Korean handwriting fonts that increase readability and completeness while maintaining the unique characteristics of handwriting, which can reduce the time and labor required to produce Korean fonts consisting of 11,172 characters. In addition, FFGAN shows effective results in correcting Chinese handwriting, so it can be used in the production of multilingual handwriting fonts.