A Fast Image Artistic Processing Method Supporting Multiple Styles

This paper presents a fast image artistic processing method supporting multiple styles, which makes the target image learn the color and brush characteristics of the multiple reference images fast under user’s control. Firstly, this method modified the anisometric texture synthesis algorithm to learn the texture and brush characteristics of the reference images by the guide of the target image’s direction field; Secondly, a parallel synthesis algorithm based on CUDA was implemented to improve the learning efficiency; Thirdly, in order to improve the controllability of the learning results, a user interactive style editing interface was designed, which achieved regional related image stylization supporting multiple styles according to the different sample style patches specified by the user, and provided WYSIWYG design experience. The experimental results show that our style learning method has faster synthesis speed and more flexibility compared with the existing stylization method based on learning.