Example-based Caricature Synthesis via Feature Deviation Matching
Example-based Caricature Synthesis via Feature Deviation Matching
复制标题
DOI：
10.1145/2949035.2949056
复制
发表时间：
2016-06
期刊：
影响因子：
--
通讯作者：
Honglin Li;Wei Yang;Hongchang Sun;M. Toyoura;Xiaoyang Mao
中科院分区：
文献类型：
--
作者：
Honglin Li;Wei Yang;Hongchang Sun;M. Toyoura;Xiaoyang Mao
生成与人脸图像相对应的漫画的技术近年来发展迅速。本文提出了一种基于特征偏差匹配的漫画合成方法。现有的方法需要建立人脸图片-漫画教学对数据库，很难收集到足够大的教学对数据库。所提出的方法解决了这个问题，通过建立对应的面部特征之间的示例漫画和输入的人脸图片的基础上，他们的偏离平均在其特征向量空间。实验结果表明，该方法具有产生与人脸图片-漫画教学对结果大致相当或更高质量的潜力。
The techniques of generating a caricature corresponding to a face picture are developing rapidly these years. In this paper, we proposed a new example-based caricature synthesis technique based on feature deviation matching. Existing methods require building databases of face picture-caricature teaching pairs, suffering from the problem that it is very difficult to collect a large enough database of such teaching pairs. The proposed method solves the problem by establishing the correspondence of facial features between the example caricatures and the input face picture based on their deviations from the average in their feature vector spaces. Experiments are performed to validate that the proposed method has the potential to produce results of roughly equivalent or higher quality to the results of using face picture-caricature teaching pairs.