Hybrid Eye Model for Photorealistic Relighting and Animation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for creating high-quality animatable and relightable 3D avatars of real people face challenges in modeling human eyes, particularly in representing diffuse and reflective surfaces, disentangling skin and eye appearance from environmental illumination, and capturing eyeball motion and skin deformation for coherent viewpoint synthesis and relighting.

Innovation Solution

A hybrid representation combining explicit mesh-based and implicit volumetric models for the eyeball surface and periocular region, using a lightweight capture system with static cameras and a hand-held camera with a co-located light source, to model the eye region's geometry and appearance effectively.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If a detailed mesh-based model is used for the eyeball surface to achieve high-fidelity rendering, then the visual quality improves, but the computational complexity and processing time increase

Engineering Contradiction:
Improvevisual qualityVSAvoidcomputational complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The system segments the eye model into distinct components (eyeball surface, iris, pupil, sclera, cornea, eyelids, eyelashes, eyebrows) and applies different modeling techniques to each. The eyeball surface uses a detailed mesh-based parametric model for high-fidelity rendering, while other components use appropriate representations (implicit surfaces for volumetric regions, geometric primitives for simple structures). This segmentation allows high visual quality where needed without applying complex modeling everywhere, thus reducing overall computational complexity.

Inventive Principle:
Principle #1Segmentation

2Manufacturing precision

If multiple static cameras and lights are used to capture the eye region from multiple viewpoints, then the synthesis quality improves, but the capture system complexity and cost increase

Engineering Contradiction:
Improvesynthesis qualityVSAvoidcapture system complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The system uses a single hand-held camera with a co-located light source that serves multiple functions: capturing images from different viewpoints, providing illumination, and enabling both geometric and appearance capture. This multi-functional device replaces what would traditionally require multiple specialized cameras and light sources, reducing capture system complexity while maintaining synthesis quality through the unified data collection approach.

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Manufacturing precision

If the eye model includes detailed representation of periocular region and interior eye volume, then the realism improves, but the data processing and storage requirements increase

Engineering Contradiction:
ImproverealismVSAvoiddata processing requirements
Core Design Contradiction:
Manufacturing precisionVSQuantity of substance

Solution Approach 1:

The system applies different levels of detail and modeling approaches to different regions of the eye model. The eyeball surface receives a detailed parametric mesh model for high realism, while the interior eye volume and periocular region use implicit volumetric representations that are more compact. This local differentiation of model quality allows realistic rendering where visually critical while using more efficient data structures elsewhere, reducing overall data processing and storage requirements.

Inventive Principle:
Principle #3Local quality

Data Source

PatentUS20240029333A1Hybrid representation for photorealistic synthesis, animation and relighting of human eyes
Publication Date: 2024.01.25 GOOGLE LLC
  • US20240029333A1 patent drawing
  • US20240029333A1 patent drawing
  • US20240029333A1 patent drawing

AI summary

A method including selecting a first point from a 3D model representing an avatar, the first point being associated with an eye, selecting a second point from the 3D model, the second point being associated with a periocular region associated with the eye, generating an albedo and spherical harmonics (SH) coefficients based on the first point and the second point, and generating an image point based on the albedo, and the SH coefficients.