Image Registration via Probability-Weighted Similarity

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current non-rigid registration techniques for images from different modalities, such as CT and MRI, face challenges in accurately evaluating similarity between pixel values, leading to incorrect transformation parameters and registration issues due to the limited range of pixel values and the reliance on correlation between distributions without considering probability information.

Innovation Solution

Incorporating probability information into the evaluation function to weight the similarity between pixel values, using conditional probability distributions to assess the reasonableness of pixel value combinations, and employing terms like mutual information or square loss mutual information to refine the similarity evaluation, ensuring accurate deformation of images to align spatial positions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If only the correlation between distributions of pixel values is used as the measure of similarity, then the evaluation process is simple, but the similarity evaluation may be incorrect for impossible pixel value combinations

Engineering Contradiction:
Improveevaluation process complexityVSAvoidsimilarity evaluation accuracy
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The patent changes the evaluation parameter from simple correlation to a probability-based measure that incorporates the reasonableness of pixel value combinations. By using probability information to weight the similarity evaluation, the system maintains computational feasibility while significantly improving accuracy in distinguishing valid from invalid pixel value pairings across modalities.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent introduces probability information as an intermediary element that mediates between the pixel values of different modalities and the similarity evaluation. This probability layer acts as a filter that weighs whether a given pixel value combination is reasonable, thereby improving the accuracy of the similarity measure without directly modifying the pixel values themselves.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If probability information is incorporated into the evaluation function, then the similarity evaluation accuracy is improved, but the computational complexity increases

Engineering Contradiction:
Improvesimilarity evaluation accuracyVSAvoidevaluation function complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies probability information selectively rather than uniformly across all pixel comparisons. By focusing the probability-based evaluation on critical regions or key pixel value ranges, the system achieves improved accuracy without the full computational burden of applying probability weighting to every possible pixel combination across the entire image space.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP2796089B1Image processing device and image processing method, and image processing program
Publication Date: 2019.05.22 FUJIFILM CORP
  • EP2796089B1 patent drawingFigure 1~2
  • EP2796089B1 patent drawingFigure 3A~3B
  • EP2796089B1 patent drawingFigure 4A~4B

AI summary

[Problem to be solved] To accurately register two images obtained by imaging the same subject with different types of modalities to make the spatial positions of the subject in the two images conform to each other. [Solution] A first image (V1) and a second image (V2) obtained by imaging the same subject with different types of modalities are obtained. The first image (V1) is deformed, and similarity between the deformed first image (V1a) and the second image (V2) is evaluated by an evaluation function that evaluates correlation between distributions of corresponding pixel values of the two images to estimate an image deformation amount of the first image (V1). Based on the estimated image deformation amount, a deformed image of the first image (V1) is generated. The evaluation function includes a term representing a measure of correlation between a pixel value of the deformed first image (V1a) and a corresponding pixel value of the second image (V2), wherein the term evaluates the correlation based on probability information that indicates a probability of each combination of corresponding pixel values of the first image (V1) and the second image (V2).