3d Face Modeling Based On Orthogonal Photos

In this amazingly civilized era which is based on advanced IT and LSI chips, 3-Dimensional Computer Graphics technology and 3D animation have been widely applied in many fields in society, and have been in the spotlight among many researchers. 3D face modeling and 3D facial animation have been successfully applied to film production, video game industry, computer-aided treatment and diagnosis, novel human-machine interactive interface etc.In this thesis on face modeling based on orthogonal images and generic 3D face model, iterative interpolation and shape-based triangular analysis and subdivision are proposed. Research work have been done in this thesis are listed below.(1)In this thesis, Candide-3 generic face model is researched in details, including how to use it, its parameters for facial animation, and its relation to MPEG-4 facial animation standards.(2)Iterative interpolation and subdivision technology is proposed and presented. After generated the 3D coordinates for the facial feature vertices in initial 3D personalized face, non-feature vertices in personalized face are synthesized using interpolation technology. Subsequently, triangular fragments in regions where visual effect and reality are unsatisfied are subdivided to produce realistic face. After subdivision, non-feature vertices are produced. Then new round of iteration of interpolation is implemented to transform these non-feature vertices into feature vertices by computing their 3D coordinates and adding them to the set of feature vertices. This process lasted until the visual effect of all regions meets demands.(3) Shape-based triangular analysis and subdivision are brought forward. Traditional subdividing algorithms tend to subdivide triangles in a unified way for all triangles but in different shape. However, this might result in extremely narrow slits in personalized face, which hinder the contribution of the subdivision to a realistic 3D face. In this thesis, the shape of the triangular fragments determined how to subdivide them and then effectively prevent the result of slits.Experimental results showed that the proposed algorithm is effective to produce a realistic 3D face model, and achieved a tradeoff between the reality of face and the computing time.