摘要
文中为人脸图片或视频的风格化生成提出一种新方法。首先,通过主动外观模型(AAM)识别人脸的轮廓及五官,记录这些特征点的位置;然后,从现有样图中提取风格化因子,通过几何变形等手段映射到任意一张输入人脸图片上,合成具有样图风格化效果的人脸图片。本方法可以与许多应用结合,如电子贺卡、网上聊天等,可实现自动生成具有表情变化的风格化视频。此方法还有易于训练、速度快、扩展性强等特点。
Presents a new method for generating stylized facial picture or video. First, using training active appearance models to tracking facial feature and contours and mark the feature points. Then, extract stylized factor from stylized sample and corresponding realistic image. Apply tiffs stylized factor into any facial face, by means of metamcrphosis according to feature points. This method can be combined into many applications such as electronic card, online chatting. Can use this method to obtain stylized facial video with expressions. In addition, this method is easy to training, fast to complement and expandable.
出处
《计算机技术与发展》
2009年第2期127-129,136,共4页
Computer Technology and Development
基金
浙江大学科研训练项目