Chinese Painting Rendering by Adaptive Style Transfer
Chinese painting is distinct from other art in that the painting elements are exhibited by complex water-and-ink diffusion and shows gray, white and black visual effect. Rendering such a water-and-ink painting with polychrome style is a challenging problem. In this paper, we propose a novel style transfer method for Chinese painting. We firstly decompose the Chinese painting with adaptive patches based on its structure, and locally colorize the painting. Then, the colorized image is used for guiding the process of texture transfer that is modeled in Markov Random Field (MRF). More precisely, we improve the classic texture transfer algorithm by modifying the compatibility functions for searching the optimal matching, with the chromatism information. The experiment results show that proposed adaptive patches can well preserve the original content while match the example style. Moreover, we present the transfer results with our method and recent style transfer algorithms, in order to make a comparison.
Chinese painting rendering Style transfer Adaptive patch-based texture transfer Markov Random Field
Wanxin Zou Xutao Li Sengping Li
Department of Electronic Engineering,Shantou University,Shantou 515063,China Department of Mechanical and Electronic Engineering,Shantou University,Shantou 515063,China
国际会议
广州
英文
3-14
2018-11-23（万方平台首次上网日期，不代表论文的发表时间）