An automated technique for the generation of facial surface models as an aid in orthodontic and facial orthognathic research

摘要

A non-invasive technique for the automatic production of surface models from stereo imagery of human faces is described. The technique uses imagery in a digital format and is designed for use on a personal workstation, such as an IBM RS6000, Acorn R260 or Sun SPARC. Once the stereo imagery has been obtained it is correlated using a coarse-to-fine area based stereo matcher. This differs from many current stereo matching algorithms in that it is designed to correlate quasi continuous surfaces, such a faces, rather than isolated point or line features. The result is the production of a dense digital disparity model (DDM) covering the stereo extent of the imagery. The DDM produced may then be projected through a numerical geometrical model of the CCD sensors used to acquire the imagery, thus yielding a digital surface model of the face. The surface model produced may then be manipulated by medical and dental practitioners to study both quantitatively and qualitatively the facial soil tissue morphology and growth and the effects of orthodontic surgery on a patient's appearance.