Image Segmentation Using Positioning and Segmentation Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image segmentation techniques are inefficient for complex images with blurred boundaries, requiring significant computing resources and resulting in poor accuracy, especially in medical imaging where clear boundary information is lacking.
Innovation Solution
A method using a positioning model and a segmentation model, both based on machine learning, to automatically identify and segment target regions in images, improving accuracy and reducing processing time by preprocessing sub-images and employing iterative training processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional image segmentation techniques are used, then the process can be implemented, but the accuracy is poor and processing time is long for complex images with blurred boundaries
Solution Approach 1:
The patent divides the image into multiple sub-images or regions of interest (ROIs) before processing. By segmenting the large image into smaller manageable parts, the system can apply segmentation models more efficiently to each sub-region, improving both accuracy for complex boundaries and overall processing speed through parallelization.
Solution Approach 2:
The patent performs preliminary processing steps including image preprocessing, feature extraction, and boundary detection before the main segmentation task. These preliminary actions prepare the image data in advance, enhancing the quality of input to the segmentation model and enabling more accurate and faster segmentation of blurred boundaries.
2Productivity
If conventional image segmentation techniques are used, then the process can be completed, but significant computing resources are required resulting in poor efficiency
Solution Approach 1:
By dividing the image into sub-images and processing them in parallel, the system reduces the computational burden on single processing units while improving overall throughput. This segmentation approach enables efficient use of computing resources through distributed processing.
Solution Approach 2:
The patent applies different processing strategies to different regions of the image based on their characteristics. Regions with clear boundaries may use simpler processing, while regions with blurred boundaries receive enhanced processing only where needed, optimizing computational resource allocation.
3Speed
If the image segmentation process is simplified, then processing speed increases, but accuracy decreases especially for blurred boundaries
Solution Approach 1:
The patent performs preliminary boundary detection and feature extraction to identify and mark regions with blurred boundaries before applying the segmentation model. This allows the system to apply enhanced processing specifically to problematic regions while maintaining fast processing for clear regions.
Solution Approach 2:
The system applies different levels of processing quality to different image regions. Regions with clear boundaries are processed quickly with standard algorithms, while regions with blurred boundaries receive enhanced processing to maintain accuracy, thus balancing speed and precision locally.
Data Source
AI summary
The present disclosure provides methods and systems for image segmentation. The methods may include obtaining an initial image. The methods may include determining, based on the initial image, at least one target image using a positioning model. Each of the at least one target image may include a portion of a target region. The methods may further include, for the each of the at least one target image, determining an initial segmentation result of the portion of the target region using a segmentation model.