Proceedings of the 28th annual conference on Computer graphics and interactive techniques
ACM SIGGRAPH 2003 Papers
Differential Coordinates for Interactive Mesh Editing
SMI '04 Proceedings of the Shape Modeling International 2004
Interactive digital photomontage
ACM SIGGRAPH 2004 Papers
Deformation transfer for triangle meshes
ACM SIGGRAPH 2004 Papers
Proceedings of the 2004 Eurographics/ACM SIGGRAPH symposium on Geometry processing
Vignette and Exposure Calibration and Compensation
ICCV '05 Proceedings of the Tenth IEEE International Conference on Computer Vision (ICCV'05) Volume 1 - Volume 01
Efficient gradient-domain compositing using quadtrees
ACM SIGGRAPH 2007 papers
On Linear Variational Surface Deformation Methods
IEEE Transactions on Visualization and Computer Graphics
Fast Construction of k-Nearest Neighbor Graphs for Point Clouds
IEEE Transactions on Visualization and Computer Graphics
Hi-index | 0.00 |
In this paper we present an interactive editing system that allows digital modelers to rapidly create new blendshape face models for incidental or background characters, starting from a small number of given face models. The obvious approach of generating linear combinations of the given models is limited in that it does not explore the combinatorial variety in the given source models. Instead, we take an approach similar to the "facial composite" systems used in forensics, in which face features (eyes, nose, mouth, etc.) from different individuals are assembled to create a composite likeness. A literal implementation of the facial composite idea does not produce realistic faces however, since choosing features from faces of different sizes can result in both unrealistic discontinuities and position relationship violations (a selected mouth might be at a higher position that the bottom of the nose selected from a different character). To handle this, we adopt a gradient-domain assembly and blending of the individual features, and produce the final face by integrating the resulting Poisson equation. This method guarantees that the result is smooth and that position relationships from the source characters are preserved. Additional blendshape expression targets are generated by the same approach, using the same gradient blending used to produce the neutral face. The resulting system is simple, operates at interactive rates, and has already proven itself in production.