Multi-Resolution Image Segmentation via Sub-Image Merging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High-resolution images in medical image processing often result in inaccurate image segmentation due to the limitations of neural networks in processing large images, leading to imprecise segmentation of objects, especially in edge regions.
Innovation Solution
The method involves acquiring two images of the same imaging target with different resolutions, where the high-resolution image is segmented into sub-images and processed in parallel with a second image of lower resolution, and the initial segmentation results are merged to enhance accuracy and precision, focusing on local and global features respectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the high-resolution image is processed directly using neural networks, then the segmentation detail accuracy is improved, but the processing time and computational resources increase significantly
Solution Approach 1:
The high-resolution image is divided into multiple sub-images, which are then processed in parallel by neural networks. This segmentation approach reduces the computational burden on each processing unit while maintaining overall segmentation accuracy, effectively resolving the contradiction between detail accuracy and processing time.
Solution Approach 2:
The patent introduces a multi-scale processing dimension by combining results from different resolution levels. Low-resolution images provide global context while high-resolution sub-images provide local details, creating a hierarchical processing structure that accelerates convergence without sacrificing precision.
2Productivity
If the image is split into small blocks for neural network processing, then the processing efficiency is improved, but the segmentation accuracy deteriorates
Solution Approach 1:
The patent merges segmentation results from multiple sources: low-resolution global segmentation results and high-resolution sub-image segmentation results. By combining these complementary information sources, the method recovers boundary details lost in low-resolution processing while maintaining the efficiency benefits of block-based processing.
Solution Approach 2:
The low-resolution image serves as an intermediary that provides global contextual information and initial segmentation guidance. This intermediary result guides the processing of high-resolution sub-images, ensuring that local details are segmented in the correct global context, thereby improving overall accuracy without sacrificing processing efficiency.
3Measurement precision
If only high-resolution images are used for segmentation, then the local feature accuracy is improved, but the global context understanding deteriorates
Solution Approach 1:
The patent applies different processing qualities to different parts of the image: low-resolution processing for global context and high-resolution processing for local features. This local quality differentiation ensures that each region is processed at the appropriate resolution level, optimizing both global understanding and local accuracy.
Solution Approach 2:
The segmentation architecture implements a nested structure where low-resolution global segmentation results are embedded within high-resolution sub-image processing. The global context from the lower-resolution level guides and constrains the local processing at higher resolutions, creating a hierarchical nested processing system that preserves both global and local information.
Data Source
AI summary
Segmenting an image is disclosed including acquiring a first image and a second image, the first image and the second image being obtained based on a same imaging target, and a resolution of the first image being greater than a resolution of the second image, performing image segmentation processing based on a plurality of sub-images of the first image to obtain a first initial segmentation result, performing image segmentation processing based on the second image to obtain a second initial segmentation result, merging, based on the imaging target, the first initial segmentation result with the second initial segmentation result to obtain a target segmentation result, and outputting or storing the target segment result.