Image Correspondence Evaluation Using Quality-Based Selection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current image processing methods for driver assistance systems face challenges in efficiently evaluating correspondences between images, leading to data redundancy and reduced accuracy due to the lack of selective and accurate correspondence evaluation.

Innovation Solution

A method that provides correspondences between images, assigns quality measures as attributes, evaluates, and conditionally selects correspondences, reducing data requirements while maintaining accuracy by using a combination of attributes for evaluation and selection.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If all correspondences between images are processed without selection, then complete data is available for evaluation, but data redundancy increases and processing efficiency decreases

Engineering Contradiction:
Improveevaluation accuracyVSAvoidprocessing efficiency
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent applies preliminary action by assigning quality measures to correspondences before the main evaluation process. The system pre-calculates attributes such as matching quality, spatial distribution, and temporal consistency for each correspondence, then uses these pre-computed values to filter and select high-quality correspondences. This preliminary classification enables efficient processing by eliminating low-quality data early, resolving the contradiction between maintaining evaluation accuracy and improving processing efficiency.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If multiple quality measures are used to evaluate correspondences, then selection accuracy improves, but computational complexity increases

Engineering Contradiction:
Improvecorrespondence selection accuracyVSAvoidevaluation system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the quality evaluation process into multiple independent attributes, each measuring a specific aspect of correspondence quality. These include matching quality (image similarity), spatial distribution (geometric consistency), temporal consistency (motion smoothness), and outlier detection (anomaly identification). By dividing the complex evaluation into modular, independent quality measures, the system achieves high selection accuracy while managing computational complexity through structured organization and selective application of different attributes based on processing needs.

Inventive Principle:
Principle #1Segmentation

3Loss of information

If all correspondences are retained for further processing, then no data is lost, but data volume and storage requirements increase

Engineering Contradiction:
Improveinformation completenessVSAvoiddata volume
Core Design Contradiction:
Loss of informationVSQuantity of substance

Solution Approach 1:

The patent applies local quality by evaluating and selecting correspondences based on their individual quality attributes rather than treating all data uniformly. The system identifies and retains only high-quality correspondences that meet specific thresholds for matching quality, spatial distribution, and temporal consistency. This selective retention maintains the essential information needed for accurate evaluation while significantly reducing the volume of data that requires storage and further processing, eliminating the need to preserve low-quality or redundant correspondences.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP3685352B1Method and device for evaluating images, operating assistance method, and operating device
Publication Date: 2024.03.06 ROBERT BOSCH GMBH
  • EP3685352B1 patent drawingFigure 1
  • EP3685352B1 patent drawingFigure 2
  • EP3685352B1 patent drawingFigure 3

AI summary

The present invention relates to a method (S) for evaluating images (B1, B2) and in particular for evaluating correspondences (10, 12) of images (B1, B2), the method comprising: (i) providing (S1-1) correspondences (10, 12) between given first and second images (B1, B2); (ii) providing (S1-2) a quality measure or a plurality of quality measures as attributes (20, 22) for characterising each correspondence (10, 12); (iii) evaluating (S2-1) and, as required, selecting (S2-2) the correspondences (10, 12); (iv) providing (S2-3) selected correspondences (10, 12) as the evaluation result (100), wherein a combination of attributes (20, 22) are taken as a basis for the evaluation (S2-1) of correspondences (10, 12), and a result of the evaluation (S2-1) is taken as the basis for the selection (S2-2) of correspondences (10, 12).