AR Face Illumination Estimation for Virtual Object Lighting

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing augmented reality (AR) technologies face challenges in achieving consistent lighting for virtual objects integrated into real environments, especially in environments with unknown and arbitrary lighting conditions.

Innovation Solution

The method involves capturing an image of a user's face with a user-facing camera, estimating the head pose, and then using a facial illumination estimator to reconstruct the face-incident light. This light is used to render virtual objects with consistent illumination, blending them seamlessly into the real environment using differential rendering.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Illumination intensity

If additional setups like mirrors or special cameras are used to estimate environment light, then lighting consistency for virtual objects is improved, but device complexity and applicability are worsened

Engineering Contradiction:
Improvelighting consistencyVSAvoidadditional setups
Core Design Contradiction:
Illumination intensityVSDevice complexity

Solution Approach 1:

The system uses the existing front-facing camera to capture images of the user's face, which already contains illumination information. By analyzing the face image directly, the system extracts light direction and intensity without requiring any additional light-capturing devices or mirrors. The face itself serves as the light probe, eliminating the need for separate measurement equipment.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The front-facing camera serves dual purposes: capturing the user's face for AR overlay positioning and simultaneously capturing illumination information for lighting estimation. This multi-functional use of the existing camera eliminates the need for dedicated light measurement devices while maintaining lighting consistency.

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

2Illumination intensity

If panoramic images or mirror sphere images are used to capture environment light, then lighting estimation is improved, but device complexity and scene alteration are worsened

Engineering Contradiction:
Improvelight estimation accuracyVSAvoidomni-camera setup
Core Design Contradiction:
Illumination intensityVSDevice complexity

Solution Approach 1:

The user's face acts as a self-contained light probe that directly reveals the illumination characteristics of the environment. By analyzing how light interacts with the face's known geometry and material properties, the system extracts accurate lighting information without needing to capture panoramic views or use mirror spheres.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system extracts only the essential illumination information (light direction, intensity, and color) from the face image, separating this critical data from the rest of the scene. This extraction approach provides sufficient lighting estimation without capturing or processing the entire environmental panorama.

Inventive Principle:
Principle #2Taking out (Extraction)

3Measurement precision

If common approaches are used to estimate environment light directions, then lighting direction is improved, but light source position information is lost

Engineering Contradiction:
Improvelight direction estimationVSAvoidlight source position
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

The system uses the known 3D geometry and material properties of the face as a reference to interpret the captured face image. By comparing the observed lighting patterns on the face with predictions from light transport models, the system iteratively refines estimates of both light direction and light source position, with each informing the other to achieve accurate results.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP3629303B1Method and system for representing a virtual object in a view of a real environment
Publication Date: 2025.05.14 APPLE INC
  • EP3629303B1 patent drawingFigure 1
  • EP3629303B1 patent drawingFigure 2
  • EP3629303B1 patent drawingFigure 3

AI summary

The invention relates to a method of representing a virtual object in a view of a real environment which comprises the steps of providing image information of a first image of at least part of a human face captured by a first camera, providing at least one human face specific characteristic, determining at least part of an image area of the face in the first image as a face region of the first image, determining at least one first light falling on the face according to the face region of the first image and the at least one human face specific characteristic, and blending in the virtual object on a display device in the view of the real environment according to the at least one first light. It further comprises providing a plurality of face sample positions, providing, for each of the plurality of face sample positions, a radiance transfer function, determining image positions of at least part of the face sample positions in the first image, and determining the at least one first light according to intensity information associated with the image positions of at least part of the face sample positions in the first image and the radiance transfer functions associated with the at least part of the face sample positions in the first image. The invention also relates to a system for representing a virtual object in a view of a real environment.