Anisotropic Multi-Scale Image Contrast Enhancement

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing multi-scale image processing methods struggle to selectively enhance image features, such as chromosome bands, while preserving edge transitions and orientation, particularly in medical images with sharp grey level transitions like CT images.

Innovation Solution

The method employs anisotropic multi-scale image processing using orientation maps to steer the enhancement of translation difference images and apply anisotropic weighting, allowing for directional enhancement of image features by computing center differences based on local orientations and applying conversion operators to translation differences before summation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Illumination intensity

If conversion functions are excessively non-linear to enhance contrast, then image contrast is improved, but grey value transitions are distorted and artifacts are created

Engineering Contradiction:
Improveimage contrastVSAvoidartifacts and distortion
Core Design Contradiction:
Illumination intensityVSObject-generated harmful factors

Solution Approach 1:

The patent segments the image processing into multiple scales using a pyramid structure, where each level processes different frequency components. This allows non-linear contrast enhancement to be applied selectively to specific frequency bands rather than the entire image, reducing artifact propagation while maintaining contrast improvement in relevant regions.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies different processing characteristics to different regions of the image based on local properties. By computing enhancement parameters locally and applying them in the frequency domain, the method preserves edge sharpness in critical regions while allowing more aggressive contrast enhancement in homogeneous areas, thus improving contrast without creating harmful artifacts.

Inventive Principle:
Principle #3Local quality

2Illumination intensity

If multi-scale techniques are applied to CT images to enhance contrast, then image contrast is improved, but edge transitions are distorted and homogeneity is lost

Engineering Contradiction:
Improveimage contrastVSAvoidedge transition shape
Core Design Contradiction:
Illumination intensityVSShape

Solution Approach 1:

The patent decomposes the image into a multi-scale pyramid where each level represents different frequency components. By processing each scale separately and using appropriate conversion functions for each level, the method enhances contrast while preserving the structural integrity of edge transitions through controlled reconstruction.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent incorporates feedback mechanisms in the reconstruction process where information from multiple scales is combined with weighting factors that preserve edge characteristics. The reconstruction algorithm uses feedback from the processed detail images to maintain original edge shapes while incorporating enhanced contrast information.

Inventive Principle:
Principle #23Feedback

3Ease of manufacture

If isotropic filters are used in decomposition and reconstruction, then processing is simplified, but selective enhancement of oriented features is not possible

Engineering Contradiction:
Improveprocessing simplicityVSAvoidselective feature enhancement capability
Core Design Contradiction:
Ease of manufactureVSAdaptability or versatility

Solution Approach 1:

The patent transitions from static isotropic filtering to dynamic anisotropic filtering where the filter characteristics adapt to local image structures. By computing orientation maps and using them to steer the filtering process, the method maintains processing efficiency while gaining the versatility to selectively enhance features in specific orientations and locations.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent changes the parameters of the filtering operation based on local image properties. By computing orientation information and using it to modulate the filter response, the method preserves the simplicity of the overall processing framework while enabling selective enhancement of oriented features through parameter adaptation.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP2198402B1Method of generating a multiscale contrast enhanced image
Publication Date: 2019.06.12 AGFA NV
  • EP2198402B1 patent drawingFigure 1
  • EP2198402B1 patent drawingFigure 2
  • EP2198402B1 patent drawingFigure 3

AI summary

At least one approximation image is created of the image at one or multiple scales. Translation difference images are created by pixel-wise subtracting the values of an approximation image at scale s and the values of a translated version of the approximation image. A non-linear modification is applied to the values of the translation difference image (s) and at least one enhanced center difference image at a specific scale is computed by combining the modified translation difference images at that scale or a smaller scale with weights Wi,j. An enhanced image is computed by applying a reconstruction algorithm to the enhanced center difference images. The non-linear modification of the values of the translation difference images is steered by the values of an orientation map which comprises for each pixel a local direction of interest. In addition or alternatively, at least one enhanced center difference image is computed by anisotropic weighing of the enhanced translation differences with weights steered by the orientation map.