Multi-Camera Image Registration for Coherent 3D Surround View

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing vehicle surround view monitoring systems face image registration issues due to mismatches between images captured by different cameras, leading to discord and reduced discernability of the vehicle's surroundings, particularly in three-dimensional displays.

Innovation Solution

An image registration apparatus and method that utilize a processor to project, map, and determine match rates between images from multiple cameras, adjusting model coordinates to achieve registration by transforming models based on preset parameters to ensure alignment and improve image coherence.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Area of stationary object

If multiple images from different cameras are combined to create a three-dimensional surround view, then the coverage of the surrounding environment is improved, but image mismatch occurs in boundary regions leading to reduced discernability

Engineering Contradiction:
Improvecoverage areaVSAvoidimage alignment accuracy
Core Design Contradiction:
Area of stationary objectVSMeasurement precision

Solution Approach 1:

The patent transforms two-dimensional images from multiple cameras into a three-dimensional coordinate system by introducing depth information (Z-axis) based on camera positions and focal lengths. This dimensional transformation enables accurate spatial mapping and registration of images from different viewpoints, resolving the boundary mismatch problem while maintaining comprehensive coverage.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Solution Approach 2:

The patent adjusts multiple parameters including camera positions, directions, focal lengths, and projection distances to achieve proper image registration. By dynamically modifying these parameters during the mapping process, the system aligns boundary regions of adjacent images while preserving the three-dimensional spatial relationships, thus eliminating discord without sacrificing coverage area.

Inventive Principle:
Principle #35Parameter changes

2Adaptability or versatility

If images are captured from different angles and positions by multiple cameras, then comprehensive surrounding view is achieved, but mismatch in boundary regions occurs reducing image quality

Engineering Contradiction:
Improvesurrounding view coverageVSAvoidimage coherence
Core Design Contradiction:
Adaptability or versatilityVSReliability

Solution Approach 1:

The patent introduces a three-dimensional coordinate system as an intermediary framework that mediates between images from different cameras. This intermediate spatial representation serves as a common reference frame, enabling consistent mapping and registration of images from various angles and positions, thus maintaining both versatility and reliability.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent creates a universal three-dimensional output model that can accommodate images from multiple cameras with different specifications (positions, angles, focal lengths). This unified framework processes and integrates diverse image sources consistently, ensuring coherent registration across all boundary regions while maintaining comprehensive surrounding view coverage.

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

Data Source

PatentEP4131164B1Method and apparatus for image registration
Publication Date: 2025.08.20 HYUNDAI MOBIS CO LTD
  • EP4131164B1 patent drawingFigure 1
  • EP4131164B1 patent drawingFigure 2
  • EP4131164B1 patent drawingFigure 3A~3B

AI summary

An image registration apparatus including at least one processor and configured to project, to a first model, a first image generated based on an image obtained from a first camera to generate a first intermediate image, to map the first intermediate image to a first output model to generate a first output image, to project, to a second model, a second image generated based on an image obtained from a second camera to generate a second intermediate image, to map the second intermediate image to a second output model to generate a second output image, and to determine a match rate between the first output image and the second output image and transform at least one of the first model and the second model based on a determined match rate and a preset reference match rate.