Synthesizing Realistic Expressions in 3D Face Data Sets

Removal of 3D facial expressions: A learning-based approach

“…4) It is helpful for face synthesis. Most of the existing methods for face synthesis work with neutral faces [11,21,33,17]. Given an expressional face, changing it to another expression is beyond the capability of the current methods.…”

“…The current 3D expression synthesis approaches can be roughly classified into four categories: 1) Interpolation-based methods define an interpolation function to specify smooth motions between two key frames [18,21]. Although interpolation usually is fast, they cannot generate arbitrary realistic facial expressions ; 2)Approaches based on physical muscle models propagate muscle strength in an elastic spring mesh to model facial expressions [6,30,10]; 3) Example-based methods synthesize expressions from a neutral face based on a collections of examples [11,17]. It can synthesize expressions in real-time with low computational cost; 4) Pseudo muscle models simulate muscle forces by splines, tensors and free deformation models [27,33,24].…”

Vertex-Based Facial Animation Transfer System for Models with Different Topology; A Tool for Maya

“…The only issue caused by the system is that it only covers the frontal face. They are focusing on making the transfer of expression for 3D sketches in the future.Minoi et.al [22]. developed a robust system that would transfer realistic expressions to a 3D face model captured by frontal photographs with a neutral face.…”

“…The 3D face model created can render in a variety of poses and illumination. "The SDM approach extracts expression discriminant information efficiently, providing a gradual transformation on the 3D faces[22]". The main advantage of the system is that the variety of expressions can be easily generated without performing intense changes in the 3D database.…”