摘要
为实现角色表情动画的情感体现,并有效去除动画边缘产生的锯齿形状的毛刺,提出基于PCNN的角色表情动画去毛刺生成方法。该方法采用PCNN网络分割采集的视频帧图像,去除图像中背景部分。将保留的部分人脸表情输入改进的深度学习模型中,提取人脸表情特征。引入风格和内容的双向约束,确定模型损失函数。融合提取的全部表情特征,生成角色表情动画。采用区域重构算法去除生成动画的边缘锯齿毛刺,保证动画生成效果。测试结果表明:图像分割性能良好,区域内部均匀性准则指标结果均在0.924以上,可清晰地保留部分人脸表情图像,能够依据真实人脸表情生成不同角色的表情动画;形状测度准则和皮尔逊相关系数指标最大值分别为0.994和0.957,可有效去除动画边缘毛刺,保证动画质量。
The method uses PCNN network to segment the captured video frame images,remove the background part of the image,and input the retained part of the face expression into the improved deep learning model to extract the face expression features.Bidirectional constraints of style and content are introduced to determine the model loss function.All the extracted expression features are fused to generate the character expression animation.The region reconstruction algorithm is used to remove the edge burrs of the generated animation to ensure the animation generation effect.The test results show that:the image segmentation performance is good,and the results of the intra-region uniformity criterion are all above 0.924,which can clearly retain part of the face expression image and generate expression animation of different characters based on the real face expression;and the maximum values of the shape measurement criterion and the Pearson’s correlation coefficient indicator are 0.994 and 0.957 respectively,which can effectively remove the edge burr of the animation to ensure the quality of the animation.
作者
范蕴睿
FAN Yunrui(School of Art and Design,Anhui Broadcasting Movie and Television College,Hefei 230011,China)
出处
《常州工学院学报》
2024年第1期27-33,共7页
Journal of Changzhou Institute of Technology
基金
安徽省教育厅高等学校省级质量工程项目(2022jyxm229)
安徽省教育厅高校哲学社会科学研究项目(2023AH052792)。
关键词
角色表情动画
去毛刺
生成方法
特征融合
图像分割
双向约束
character expression animation
deburring
generation method
feature fusion
image segmentation
bidirectional constraint
