基于人脸关键点的表情实时动态迁移
|
摘要
为了解决现有的二维人脸表情迁移,没有实时、流畅的动态体现的问题,提出一种基于人脸关键点的表情实时动态迁移的方法,区别于现有的表情迁移的方法,使用dlib人脸识别与关键点检测,实时采集视频中人脸关键点信息,加入帧间平滑处理,使得目标人脸动态变化更流畅和真实。在贴图时使用Delaunay三角剖分,以提高贴图的表现效果。输入源摄像头获取的人脸,将表情实时迁移到目标人脸上,使目标人脸实时高效模仿源摄像头的输入人脸表情,实现人脸表情神态和面部动作的动态迁移。
In order to solve the problem of existing two-dimensional facial expression migration without real-time and smooth dynamic expression,proposes a real-time dynamic expression migration method based on key points of face, which is different from the existing expression migration method. dlib face recognition and key point detection are used to capture the key points of face in real-time video. Point information is added to the inter frame smooth processing, which makes the dynamic change of the target face more smooth and real. The Delaunay triangulation is used in the texture map to improve the performance of the map. The human face captured by the input source camera is transferred to the target face in real time, which makes the target face imitate the input facial expression of the source camera in real time and efficiently, and realizes the dynamic migration of facial expression and facial action.
In order to solve the problem of existing two-dimensional facial expression migration without real-time and smooth dynamic expression,proposes a real-time dynamic expression migration method based on key points of face, which is different from the existing expression migration method. dlib face recognition and key point detection are used to capture the key points of face in real-time video. Point information is added to the inter frame smooth processing, which makes the dynamic change of the target face more smooth and real. The Delaunay triangulation is used in the texture map to improve the performance of the map. The human face captured by the input source camera is transferred to the target face in real time, which makes the target face imitate the input facial expression of the source camera in real time and efficiently, and realizes the dynamic migration of facial expression and facial action.
引文
[1]赵丽红,刘纪红,徐心和.人脸检测方法综述[J].计算机应用研究,2004, 21(9):1-4.
[2]张会森,王映辉.人脸识别技术[J].计算机工程与设计, 2006, 27(11):19-24.
[3]李雪,朱欣娟.二维人脸表情仿真技术研究[J].纺织高校基础科学学报,2014(3):385-391.
[4]Thies J,Stamminger M,Theobalt C. Demo of Face2Face:Real-Time Face Capture and Reenactment of RGB Videos[C]. IEEE Conference on Computer Vision&Pattern Recognition. IEEE Computer Society,2016.
[5]King D E. dlib-ml:A Machine Learning Toolkit[M]. JMLR.org,2009.
[6]Bottino A G,Cumani S. Robust Identification of Face Landmarks in Profile Images[C]. Proceeding of CSCC Conference on Circuits,2008.
[7]Chew L P. Constrained Delaunay Triangulations[J]. Algorithmica,1989,4(1-4):97-108.
[8]吴莉莉. Delaunay三角剖分的几种算法综述[J].科技信息,2011(28):119-120.
[2]张会森,王映辉.人脸识别技术[J].计算机工程与设计, 2006, 27(11):19-24.
[3]李雪,朱欣娟.二维人脸表情仿真技术研究[J].纺织高校基础科学学报,2014(3):385-391.
[4]Thies J,Stamminger M,Theobalt C. Demo of Face2Face:Real-Time Face Capture and Reenactment of RGB Videos[C]. IEEE Conference on Computer Vision&Pattern Recognition. IEEE Computer Society,2016.
[5]King D E. dlib-ml:A Machine Learning Toolkit[M]. JMLR.org,2009.
[6]Bottino A G,Cumani S. Robust Identification of Face Landmarks in Profile Images[C]. Proceeding of CSCC Conference on Circuits,2008.
[7]Chew L P. Constrained Delaunay Triangulations[J]. Algorithmica,1989,4(1-4):97-108.
[8]吴莉莉. Delaunay三角剖分的几种算法综述[J].科技信息,2011(28):119-120.