Image Segmentation Using High-Order Energy Minimization

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image processing methods fail to effectively utilize high-order energy for image segmentation, particularly in distinguishing between blood vessels such as veins and arteries, leading to misclassification due to the limitations of second-order energy minimization techniques.

Innovation Solution

An image processing apparatus and method that utilizes high-order energy, specifically the max-flow-min-cut algorithm, to select pixels representing shapes like straight lines, curved lines, planes, and spheres, allowing for precise labeling and separation of images by minimizing energy functions of the Nth order or greater, where N pixels are variables, ensuring that pixels belonging to the same class are grouped together.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If second-order energy minimization is used for image segmentation, then the computation is tractable and can be solved using graph cut algorithms, but the segmentation accuracy is insufficient for distinguishing complex structures like parallel blood vessels

Engineering Contradiction:
Improvesegmentation accuracyVSAvoidenergy function order
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent changes the parameter of energy function order from second-order to fourth-order or higher to improve segmentation accuracy. This parameter change enables the energy function to capture higher-order spatial relationships and contextual information, which is crucial for distinguishing complex structures like parallel blood vessels while maintaining computational tractability through the proposed optimization approach

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If high-order energy functions (fourth-order or greater) are used for image segmentation, then the segmentation accuracy improves for complex structures, but the computational complexity increases significantly

Engineering Contradiction:
Improvesegmentation accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSPower

Solution Approach 1:

The patent segments the high-order energy minimization problem into two parts: (1) constructing a graph representation where nodes represent pixel configurations and edges represent transitions, and (2) applying graph cut algorithms to find the minimum cut. This segmentation of the computational approach makes the fourth-order or higher energy minimization tractable by leveraging the efficiency of graph-based optimization methods

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces an intermediary graph structure that mediates between the high-order energy function and the optimization algorithm. The graph serves as an intermediary representation that encodes the high-order energy constraints, allowing the use of efficient graph cut algorithms to solve what would otherwise be a computationally intractable optimization problem

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP2902970B1Image processing device, image processing program, and operating method for image processing device
Publication Date: 2018.10.24 FUJIFILM CORP
  • EP2902970B1 patent drawingFigure 1~2
  • EP2902970B1 patent drawingFigure 3~5
  • EP2902970B1 patent drawingFigure 6~8

AI summary

[Objective] To propose a method for using high order energy that is effective when using a graph cut technique for the segmentation of images. [Constitution] When using a graph cut process for binary labeling, labeling means (13) selects N (>3) pixels in image data in such a manner to represent a predetermined shape in the image, minimize the high-order energy of the Nth order or greater in which the pixel values of the N pixels are variables, and performs labeling.