Viewing Real-World Faces in 3D

Viewing Real-World Faces in 3D

Tal Hassner

The Open University of Israel, Israel

hassner@openu.ac.il

用一个reference face来变形为一组图像序列的外形，能产生这组序列的3Dshape

Pose adjustment

从USF Human ID 3-D Database中随机选择一个人脸作为reference model，并采用【33】得方法来检测68个关键点，

1.用Viola-Jones detector 来对序列图像进行人脸检测

2.将检测到的人脸扩大2.2倍，在crop,sacle到225*225

3.用【33】在2上面检测68个关键点

4.将2d图像上检测到的关键点与3dmodel上的点对应起来

5.用【8】求解代表转换的一些本质变换，得到三个参数：camera matrix3D rotation matrix，and 3D translation vector

6.relate 3D reference points with the query photo’s viewpoint by
7.These camera parameters allow us to re-render the reference model, along with its depth, in a query-adjusted pose 。

Following the pose adjustment step, the reference photo and depth, IR and DR, are in approximately the same pose as the query

Depth optimization