View-Invariant Method for Calculating 2D Optical Strain

摘要

Two-dimensional optical strain maps have been shown to be a useful feature that describes a bio-mechanical property of facial skin tissue during the non-rigid motion that occurs during facial expressions. In this paper, we propose a method for accurately estimating and modeling the three-dimensional strain impacted onto the face and demonstrate its robustness at different depth resolutions and views. Experimental results are given for a publically available dataset that contains high depth resolutions of facial expressions, as well as a new dataset collected using the Microsoft Kinect synchronized with two HD webcams.