Mesh-Driven Virtual Strands for Realistic Facial Animation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In computer-generated animations, particularly in video games, hair features like beards and eyebrows often appear unrealistic when animated using conventional single-bone skinning methods, as they can disappear inside the face or float detached, due to fixed offsets from underlying bones, rather than moving naturally with facial deformations.
Innovation Solution
Associating each strand of a hair feature with specific points on a 3D mesh, allowing the strands to be driven by the mesh's transformations rather than individual bones, thereby ensuring realistic movement and positioning in accordance with facial animations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If strands are skinned to bones using fixed offsets, then the animation process is simple and computationally efficient, but the strands appear unrealistic and may disappear inside the face or float detached
Solution Approach 1:
The patent introduces a mesh as an intermediary between bones and strands. Instead of directly skinning strands to bones, the mesh serves as a mediator that deforms with bone transformations and provides accurate surface positioning for strands. This resolves the contradiction by maintaining computational efficiency through bone-driven mesh deformation while achieving realistic strand positioning through mesh-based strand association.
Solution Approach 2:
The patent segments the animation system into three distinct levels: bones for global transformation, mesh for surface deformation and positioning, and strands for detailed hair representation. This segmentation allows each component to operate at its optimal complexity level, with bones handling coarse movement, mesh handling surface accuracy, and strands handling detailed positioning, thereby achieving realism without overwhelming computational complexity.
2Reliability
If strands are authored at fixed offsets from bones, then the authoring process is simple, but the strands do not move naturally with facial deformations
Solution Approach 1:
The patent performs preliminary association of strands with mesh surfaces during the authoring phase. By pre-computing intersection points between strands and the mesh, the system prepares strand positioning data in advance. This preliminary action enables natural strand movement during animation without requiring complex real-time calculations, thus maintaining ease of authoring while achieving natural movement.
Solution Approach 2:
The patent uses the mesh surface as a template or copy reference for strand positioning. Instead of manually positioning each strand, the system copies positioning information from the mesh surface at intersection points. This copying approach simplifies authoring while ensuring strands are correctly positioned relative to the facial surface, and it enables natural movement as the mesh deforms with facial animations.
3Manufacturing precision
If strands are associated with mesh surfaces, then computational overhead increases, but strand positioning accuracy and realism improve significantly
Solution Approach 1:
The patent implements a dynamic strand positioning system where strand associations with mesh surfaces are computed once during authoring or mesh updates, and then reused during animation. This dynamic approach computes precise positioning data when needed (at authoring or mesh update stages) and reuses it during playback, achieving high positioning precision while minimizing continuous computational energy consumption during animation.
Solution Approach 2:
The patent performs the computationally intensive strand-mesh association calculations in advance during the authoring phase or when the mesh geometry changes. By pre-computing intersection points and associations before animation playback, the system achieves high positioning precision without requiring continuous heavy computation during animation, thus reducing real-time energy consumption while maintaining accuracy.
Data Source
AI summary
A computer implemented method comprises receiving a mesh representing a 3D object, the mesh comprising a plurality of 2D shapes, receiving a strand feature comprising a plurality of virtual strands, and associating each strand of the plurality of virtual strands with a 2D shape of the mesh.