3D Convolutional Neural Network for High-Precision Image Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image segmentation technologies using traditional 2D and 3D convolutional neural networks suffer from low precision in segmenting images.

Innovation Solution

The proposed solution involves a 3D convolutional neural network model that includes an extraction module, a pixel-level saliency enhancement module, a channel-level saliency enhancement module, and a 3D residual deconvolution module. This model processes images by extracting feature maps, adjusting pixel-level parameters, enhancing channel-level features, and calculating weighting matrices to achieve high-precision segmentation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional 2D convolutional neural network is used for image segmentation, then the computational complexity is low, but the segmentation precision is low

Engineering Contradiction:
Improvesegmentation precisionVSAvoidnetwork complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent transitions from traditional 2D convolutional neural networks to 3D convolutional neural networks by adding the depth dimension (z-axis) to the conventional height and width dimensions. This dimensional extension enables the network to capture spatial relationships in three dimensions, significantly improving segmentation precision while managing computational complexity through optimized 3D convolution operations and residual connections.

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

2Measurement precision

If 3D convolutional neural network is used for image segmentation, then the segmentation precision is improved, but the computational complexity increases

Engineering Contradiction:
Improvesegmentation precisionVSAvoidcomputational energy
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent divides the 3D convolutional neural network into distinct functional modules: a 3D encoding module for feature extraction, a pixel-level saliency enhancement module for local refinement, and a channel-level saliency enhancement module for feature channel optimization. This modular segmentation allows each component to perform specialized operations, improving overall segmentation precision while reducing redundant computations and energy consumption compared to a monolithic network architecture.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP4012652B1Image segmentation method and apparatus
Publication Date: 2025.03.05 LANGCHAO ELECTRONIC INFORMATION IND CO LTD
  • EP4012652B1 patent drawingFigure 1
  • EP4012652B1 patent drawingFigure 2
  • EP4012652B1 patent drawingFigure 3~4

AI summary

Provided in the present invention are an image segmentation method and apparatus. A segmented image of each target object, after parameter adjustment, in an acquired image to be segmented is determined by using a preset 3D convolutional neural network model, specifically the process comprising : extracting, by using an extraction module in the 3D convolutional neural network model, a first feature map matrix of at least one target object of the image to be segmented; adjusting, by using a pixel-level significance enhancement module, parameters of the first feature map matrix of each target object, and determining a pixel-level weighting matrix of each target object; enhancing a matrix channel of the first feature map matrix of each target object according to a channel-level significance enhancement module, and determining a channel-level weighting matrix of each target object; performing, by using a 3D residual deconvolutional module, reduction processing on the size of a target matrix of the target object, wherein the size is obtained by increasing the sum of the pixel-level weighting matrix and channel-level weighting matrix of each target object; and determining a segmented image of each target object after parameter adjustment. On the basis of the present invention, a high-precision segmented image can be obtained.