参数化的肖像漫画生成算法

Generation algorithm of parameterized caricature

现有计算机生成漫画的方法大多基于高层语义描述和图像模板的方式来实现,主要通过基于图像变形的算法达到漫画夸张的效果,更多关注的是夸张点的位置模板和夸张的比例,很少从艺术家的绘画本质,即线条来描述夸张的尺度。参数化的方法从艺术家绘画的本质线条出发,分析夸张的基本原理,构建组成人脸夸张各器官的线条元素,将漫画的夸张特征转换为一些可控制的参数,并通过局部参数的控制调整,实现不同艺术风格的漫画。实验证明该方法实现的漫画效果线条造型丰富,夸张效果突出,较好地还原了艺术家漫画夸张的线条表现本质。

The existing methods of computer generated cartoons are mostly realized by high-level semantic analyze or template based way, those methods are more focus on exaggerated templates and proportion, but less form the essence of artist＇s pain- tings, which used the original lines to describe the art of exaggeration. This paper firstly used the parameterized method, from the essence line of artist＇s painting, constructed the lines of composition exaggeration, it transformed the exaggeration features by some controllable parameters and realized stylized caricature by adjusting parameters. The final result of caricatures genera- ted by the proposed method restore the artist＇s exaggerate style in nature.