AR Image Pose Estimation Without SLAM Point Reconstruction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional methods for positioning Augmented Reality (AR) models in a three-dimensional space, such as SLAM and VIO, require high precision inertial measurement units and involve complex point cloud reconstruction, necessitating device translation for initialization, which limits computational efficiency and application precision.

Innovation Solution

An AR device calculates homography matrices between target planes using a target matching algorithm and IMU attitude to determine displacement, eliminating the need for device translation and reducing computational complexity by avoiding three-dimensional point reconstruction.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If SLAM or VIO methods are used for AR model positioning, then positioning accuracy can be achieved, but computational complexity increases due to point cloud reconstruction requirements

Engineering Contradiction:
Improvepositioning accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent extracts only the essential homography matrix information needed for positioning from the complex SLAM/VIO processing pipeline. By calculating homography matrices between current and reference image frames using feature point matching, the system obtains sufficient positioning data without performing full point cloud reconstruction, thereby reducing computational complexity while maintaining positioning accuracy

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

Instead of using complex forward methods (SLAM/VIO with point cloud reconstruction) to achieve positioning, the patent inverts the approach by using homography matrix decomposition to directly obtain camera pose information. This reverse engineering of the positioning process eliminates unnecessary computational steps while preserving the essential positioning functionality

Inventive Principle:
Principle #13The other way round (Inversion)

2Reliability

If device translation is required for initialization, then positioning can be established, but productivity decreases due to additional operational steps

Engineering Contradiction:
Improvepositioning establishmentVSAvoidinitialization efficiency
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The system performs self-initialization by automatically selecting reference image frames and calculating homography matrices without requiring manual device translation operations. The algorithm autonomously establishes positioning by processing available image data, eliminating the need for user-initiated translation movements while ensuring reliable positioning establishment

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent performs preliminary selection and processing of reference image frames before actual positioning is needed. By pre-identifying suitable reference frames and pre-calculating their homography relationships, the system prepares positioning data in advance, eliminating the need for real-time device translation during initialization and improving overall initialization efficiency

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4105766B1Image display method, computer device and storage medium
Publication Date: 2026.02.11 TENCENT TECHNOLOGY (SHENZHEN) CO LTD
  • EP4105766B1 patent drawingFigure 1~2
  • EP4105766B1 patent drawingFigure 3
  • EP4105766B1 patent drawingFigure 4~5

AI summary

An image display method and apparatus, and a computer device and a storage medium, which belong to the technical field of intelligent terminals. The method is executed by an augmented reality (AR) device, and comprises: according to a first target plane and a second target plane, acquiring a homography matrix between the first target plane and the second target plane (201); according to the homography matrix and the attitude of an inertial measurement unit, acquiring a target displacement (202); according to the target displacement, acquiring a target pose, wherein the target pose is the position and attitude of a camera coordinate system of the current frame image in a world coordinate system (203); and according to the target pose, displaying an AR image (204). By means of the method, the step of translating an AR device to perform initialization is not required, and the AR device is also prevented from reconstructing a large number of points in a three-dimensional space, such that the amount of calculation of the AR device is reduced, and the efficiency of locating an AR model in the AR device and the effect of displaying an AR model image are improved.