3D Image Segmentation via Local Feature Correspondences

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional methods for 3D image segmentation require significant computational resources, making them inefficient for real-time processing on devices with limited capabilities, such as small 3D devices, and fail to effectively adapt 3D video content to different viewing contexts.

Innovation Solution

The method segments 3D image data by determining local features in multiple views and merging corresponding depth regions based on spatial displacement and correlation values, reducing computational load through the use of local features instead of individual pixel analysis.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional methods for 3D image segmentation are used, then segmentation accuracy is maintained, but computational resource consumption increases significantly

Engineering Contradiction:
Improvesegmentation accuracyVSAvoidcomputational resource consumption
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent applies segmentation by dividing the 3D image data into multiple views and processing local features within each view separately. This allows the system to maintain segmentation accuracy while reducing overall computational load by working with smaller, divided portions of the image data simultaneously.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extracts and processes only local features from the 3D image data rather than analyzing all pixels. By taking out and focusing on relevant local features, the system maintains segmentation quality while significantly reducing computational resource consumption.

Inventive Principle:
Principle #2Taking out (Extraction)

2Measurement precision

If conventional 3D image segmentation methods are used, then depth region identification is achieved, but processing speed decreases due to high computational requirements

Engineering Contradiction:
Improvedepth region identification accuracyVSAvoidprocessing speed
Core Design Contradiction:
Measurement precisionVSSpeed

Solution Approach 1:

The patent segments 3D image data into multiple views and processes local features in parallel across views. This segmentation approach enables real-time processing speed while maintaining accurate depth region identification by distributing computational tasks across multiple processing streams.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs partial action by processing only local features rather than complete pixel-by-pixel analysis. This partial processing approach achieves sufficient depth region identification accuracy at speeds suitable for real-time applications.

Inventive Principle:
Principle #16Partial or excessive action

3Use of energy by moving object

If local features are used instead of individual pixel analysis, then computational load is reduced, but segmentation detail may be lost

Engineering Contradiction:
Improvecomputational loadVSAvoidsegmentation detail
Core Design Contradiction:
Use of energy by moving objectVSMeasurement precision

Solution Approach 1:

The patent segments 3D image data into multiple views and processes local features within each view. By maintaining multiple view segments and processing local features in each segment, the system reduces computational load while preserving segmentation detail through the combined information from all views.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent merges corresponding depth regions across multiple views based on spatial displacement and correlation values. This merging process combines information from local features across views to maintain segmentation detail while working with computationally efficient local feature representations.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP2932710B1Method and apparatus for segmentation of 3D image data
Publication Date: 2019.02.20 HUAWEI TECH CO LTD
  • EP2932710B1 patent drawingFigure 1~2
  • EP2932710B1 patent drawingFigure 3
  • EP2932710B1 patent drawingFigure 4

AI summary

The present invention provides a method and an apparatus for real time object segmentation of 3D image data based on local feature correspondences between a plurality of views. In order to reduce the computational effort of object segmentation of 3D image data, the segmentation process is performed based on correspondences relating to local features of the image data and a depth map. In this way, computational effort can be significantly reduced and the image segmentation can be carried out very fast.