Robust Model Driven Matching Method for Face Analysis with Multi Image Photogrammetry

A vision based method to recover human faces from video sequences is presented. Although video sequences acquired from multiple static synchronized CCD cameras have been used as a tool for 3-D reconstruction of the face before, the precision and reliability remain as concerning issues, which are addressed in this paper Moreover, the presented matching algorithm is invariant to scaling, rotation and insufficient calibration. A geometric primitive guides the matching approach through object space and therefore allows to compare gray values in an unique coordinate system