计算机肖像漫画生成系统分析.docx

Abstract
As a typical art style, portrait caricature uses simple but exaggerated artistic method to describe facial features, integrating into artist’s personal painting skill as well as subjective ,it is very difficult for puter to learn and use the painting skill to create painting skill can not be to generate caricatures with an artist’s drawing style puter automatically is a hot topic in the field puter graphics, full of interest and challenges.
In this paper, the research is studied from the image template based caricature generation to non-photorealistic ,this paper focuses on the facial shape features and facial relation features, which have great impact on generating caricature:The EDFM rule is adopted to exaggerate facial shape statistical learning method is proposed to capture an artist’s painting techniques and drawing bined with correlation ,two exaggerate methods are utilized to design a caricature system which has a wider range of applications and richer exaggerate on the process of caricature generation, we contribute to addressing several key technical problems puter caricature generation.
This paper is based on the process framework of puter caricature prototype key technologies are main work and achievements are summarized as follows: In the part of facial feature extraction, build our facial model consisted of 105 feature points by referring to Candide-3 facial improved active shape model algorithm is adopted to extract facial feature points;In the part of Exaggeration,for the disadvantages such as unreality and dependence on the number of samples,a new method based on correlation is proposed to lower the dependence and improve reality and lifelikeness;In the part of image deformation,thin plate spline algorithm and grid-based deformation are adopted in this the same time, the user interface is pro