Automatic Elementary Pattern Calculation in Periodic Digital Images

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for detecting defects and characterizing dimensional variability in digital images with periodic patterns are computationally complex and require prior knowledge, making them inefficient and operator-dependent, especially when dealing with noisy images from microscopy.

Innovation Solution

A method that automatically calculates an elementary pattern in digital images using a calculation processor to determine periods of periodic repetition, project to an orthonormal reference frame, and interpolate to find an elementary reference pattern, allowing for self-adaptive defect detection and variability analysis without requiring parameter adjustments.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If template matching or Fourier transform methods are used for defect detection, then measurement precision is improved, but device complexity increases and prior knowledge is required

Engineering Contradiction:
Improvedefect detection precisionVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system automatically determines the elementary pattern and periodicity parameters from the input image itself through autocorrelation analysis, eliminating the need for external prior knowledge or manual parameter specification. The algorithm self-adapts to the image content by identifying the fundamental repeating unit and using it as the basis for defect detection.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The method performs preliminary autocorrelation analysis to identify the elementary pattern and periodicity before proceeding to defect detection. This preliminary characterization of the image structure enables subsequent defect detection to be performed more efficiently without requiring complex preprocessing or parameter specification.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If template matching or Fourier transform methods are used, then measurement precision is improved, but ease of operation deteriorates due to requiring expert knowledge

Engineering Contradiction:
Improvedefect detection precisionVSAvoidoperator independence
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The system automatically determines the elementary pattern and periodicity parameters from the input image itself through autocorrelation analysis, eliminating the need for external prior knowledge or manual parameter specification. The algorithm self-adapts to the image content by identifying the fundamental repeating unit and using it as the basis for defect detection.

Inventive Principle:
Principle #25Self-service

3Measurement precision

If conventional methods are used for noisy microscopy images, then measurement precision deteriorates, but device complexity remains high

Engineering Contradiction:
Improvedefect detection precisionVSAvoidimage noise
Core Design Contradiction:
Measurement precisionVSObject-affected harmful factors

Solution Approach 1:

The method extracts the essential periodic structure and elementary pattern from the noisy image through autocorrelation analysis, separating the signal (periodic pattern) from the noise. By focusing on the fundamental repeating unit and its mathematical properties rather than attempting to process the entire noisy image directly, the method achieves robust defect detection even in noisy microscopy images.

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP4339889A1Method and device for automatically calculating elementary pattern in a digital image representative of physical quantities comprising a periodic repetition of patterns
Publication Date: 2024.03.20 COMMISSARIAT A LENERGIE ATOMIQUE ET AUX ENERGIES ALTERNATIVES
  • EP4339889A1 patent drawingFigure 1
  • EP4339889A1 patent drawingFigure 2
  • EP4339889A1 patent drawingFigure 3

AI summary

This process includes steps implemented by a computing processor, of: - obtaining (30) an initial digital image, - determining (32) at least two periods of periodic repetition of patterns in the initial digital image, each period being defined by a basis vector having an associated direction and length, the basis vectors forming an initial reference frame, - determining (42) a projection matrix allowing passage from the initial reference frame to an orthonormal reference frame, - calculating (44) a projected reference pattern, in the orthonormal reference frame, using the projection matrix to match a plurality of pixels of the initial digital image and a pixel of said projected reference pattern, - calculating (46) an elementary reference pattern, in the initial reference frame, from the projected reference pattern in the orthonormal reference frame.