Design of Hairy Brush Characters Using Convolutional Encoder-Decoder Network

抄録

This paper considers a problem of constructing the so-called dynamic-font based characters with natural ink effects yielding aesthetic properties. For solving such a problem, we here develop a scheme for constructing hairy brush characters by using the convolutional encoder-decoder network. The dataset for training the network consists of a set of paired data of handwriting Japanese character using a black marker and the corresponding calligraphy character using a hairy brush. The performance is demonstrated by experimental studies.