漫画风格的人脸肖像生成算法

为模拟艺术家画出的具有夸张效果的肖像漫画,由用户输入正面人脸照片,通过交互获取关键特征点,根据特征点计算面部特征值来判断需要夸张的部分和各自的夸张方式;然后根据特征点自动生成不同层级的网格并分层实施夸张变形;最后进行图像处理以获得不同艺术风格的图像.该算法从研究漫画家的作画风格入手,总结了肖像漫画特定的夸张规律,所生成的漫画风格人脸肖像效果较好,能够应用于非真实感图形学和数字娱乐等领域.

A Generation Algorithm of Caricature Portrait

This paper introduces a new facial caricaturing method for 2D cartoon portraits with exaggeration to simulate the works of the artists. By the method, firstly, a photo of front face inputs and its characteristic points are specified by users. According to the characteristic values calculated based on the characteristic points, the proper parts of the face demanding deformation and their respective ways of exaggeration are decided. Then the mesh models of different layers are automatically generated based on the characteristic points and exaggerative deformations are carried out layer by layer. At last, the image is manipulated to get different styles. The study of the arithmetic starts from the research on the styles of the artist works, and summarizes the exaggeration rules of cartoon portraits. This approach can help generate relatively good cartoon portraits and can be used in unrealistic computer graphics and digital entertainment.