Wavelet Ridge Feature Superposition for Retinal Texture Analysis

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image processing techniques, particularly those using wavelet transforms, are not robust enough to handle fragmented textural features and have limitations in resolution and resolving power, making them ineffective for analyzing diffuse, irregular, or spaced patterns in images.

Innovation Solution

The method involves mapping an image along one-dimensional slices, computing wavelet scalograms, and extracting ridge features, which are then superimposed and used to derive textural information, providing a more robust analysis by focusing on fragmented features through thresholding and histogram analysis.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If wavelet transforms are used for texture modeling, then time-frequency representation and localization are improved, but robustness to fragmented textural features deteriorates

Engineering Contradiction:
Improvetime-frequency localizationVSAvoidrobustness to fragmentation
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The image is divided into multiple overlapping patches, and each patch is processed independently through the wavelet transform pipeline. This segmentation allows the method to handle fragmented features by analyzing local regions separately, then combining results to achieve robust global texture analysis that overcomes the limitations of applying wavelet transforms to the entire image at once.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extends traditional 2D wavelet analysis by incorporating a third dimension through the processing of multiple overlapping patches. This dimensional extension allows the method to capture both local and global texture characteristics, improving robustness to fragmentation while maintaining time-frequency localization capabilities.

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

2Device complexity

If traditional texture analysis methods are used, then processing simplicity is maintained, but resolution and resolving power deteriorate

Engineering Contradiction:
Improveprocessing simplicityVSAvoidresolution and resolving power
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

By segmenting the image into overlapping patches and processing each independently, the method achieves higher resolution texture analysis without requiring excessively complex global processing. The segmentation approach distributes the computational burden while improving the resolving power for detecting fine textural details and fragmented features.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies preliminary filtering and patch extraction before the main wavelet transform analysis. This preliminary action prepares the data in a form that enhances subsequent processing efficiency and resolution, allowing the system to achieve high measuring precision without proportionally increasing overall system complexity.

Inventive Principle:
Principle #10Preliminary action

3Loss of information

If comprehensive image processing is applied, then textural analysis completeness is improved, but processing time and computational load increase

Engineering Contradiction:
Improvetextural analysis completenessVSAvoidprocessing time
Core Design Contradiction:
Loss of informationVSLoss of time

Solution Approach 1:

The patent processes multiple overlapping patches, which means some regions are analyzed more than once. This excessive action ensures complete textural coverage and captures fragmented features that might be missed in single-pass processing, while the overlapping strategy is optimized to minimize redundant computation and manage processing time effectively.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP2856429B1Methods and apparatus for image processing, and laser scanning ophthalmoscope having an image processing apparatus
Publication Date: 2020.07.29 OPTOS PLC
  • EP2856429B1 patent drawingFigure 1
  • EP2856429B1 patent drawingFigure 2
  • EP2856429B1 patent drawingFigure 3

AI summary

A laser scanning ophthalmoscope obtains images of a retina. An image is processed by (i) mapping an image along a one dimensional slice; (ii) computing a wavelet scalogram of the slice; (iii) mapping ridge features from the wavelet scalogram; repeating steps (i), (ii) and (iii) for one or more mapped image slices. The mapped ridge features from the slices are superimposed. Textural information is derived from the superimposed mapped ridge features. The analysis can be tuned to detect various textural features, for example to detect image artefacts, or for retinal pathology classification.