Decoration of Human Face in Video Image by Estimating Eye Positions with Particle Filter
Decoration of Human Face in Video Image by Estimating Eye Positions with Particle Filter
复制标题
DOI：
--
复制
发表时间：
2011
期刊：
影响因子：
--
通讯作者：
Gefan Zhang;N. Ikoma;H. Kawano;H. Maeda
中科院分区：
文献类型：
--
作者：
Gefan Zhang;N. Ikoma;H. Kawano;H. Maeda
提出了一种基于粒子滤波的人眼位置估计方法来装饰视频图像中的人脸。眼睛位置的估计是在一个连续的方式，通过制定一个面部运动的状态空间模型。该模型由简化的二维人脸与二维运动和旋转。双眼在图像帧中的位置在人脸姿态估计中起着关键作用。利用粒子滤波算法，结合人脸和眼睛的适当性计算，实现了眼睛位置的实时估计。眼睛装饰的结果是通过估计瞳孔位置来扩大眼睛的面积。在静态场景、旋转、平行运动以及其他运动情况下的实验结果表明了该方法的有效性。
Decorating human face in video image by estimating eye positions with particle filter has been proposed. Estimation of eye positions is presented in a sequential manner by formulating a face motion in a state space model. The model consists of simplified twodimensional face with two dimensional motion and rotation. The positions of the two eyes in an image frame play a key role in estimating the face posture. Real-time estimation of eye positions has been realized by particle filter algorithm with likelihood computation consisting of adequacy of face and eyes. Eye decoration works as a result for enlarging the areas of eyes by estimating pupil positions. Experiments with different conditions of face, such as static scene, rotation, parallel motion, and some other motions have conducted to demonstrate performance of the proposed method.