Hierarchical Image Segmentation for Boundary Precision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image segmentation techniques require high processing costs and often result in incorrect area extraction due to processing the entire image, leading to inefficiencies and inaccuracies.
Innovation Solution
An image processing apparatus that segments an image in a hierarchical manner, starting with low resolution and using the results to limit the processing target area for higher resolution images, thereby reducing erroneous extractions and increasing accuracy and speed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the entire image is processed for segmentation, then comprehensive area extraction is achieved, but processing time and computational cost increase significantly
Solution Approach 1:
The patent applies segmentation by dividing the image processing into multiple resolution levels (coarse to fine) and further dividing the processing scope at each level by focusing only on boundary regions. This multi-level segmentation approach reduces the total processing area while maintaining comprehensive coverage, resolving the contradiction between complete segmentation and processing efficiency.
Solution Approach 2:
The patent performs preliminary segmentation at low resolution before high-resolution processing. The low-resolution segmentation results are used to pre-identify boundary regions, which then serve as the focused processing target for high-resolution segmentation. This preliminary action eliminates the need to process the entire high-resolution image, reducing processing time while maintaining accuracy.
2Reliability
If the entire image is processed for segmentation, then all potential areas are covered, but computational load and processing costs increase
Solution Approach 1:
The patent applies local quality by assigning different processing qualities to different regions of the image. Boundary regions identified from low-resolution processing receive high-resolution processing, while non-boundary regions receive lower-resolution processing or are excluded from further processing. This differential quality approach maintains extraction reliability for important regions while reducing overall computational cost.
Solution Approach 2:
The patent uses partial action by processing only the necessary portions of the image at high resolution. Instead of uniformly processing the entire image, the method selectively applies high-resolution segmentation only to boundary regions where accurate extraction is critical, achieving reliable results with reduced computational expenditure.
3Productivity
If low resolution segmentation is performed first, then processing speed is improved, but the accuracy of area extraction may decrease
Solution Approach 1:
The patent resolves the accuracy concern by adding a dimensional aspect to the processing pipeline. After low-resolution segmentation provides initial boundary identification, the method transitions to a higher resolution dimension for processing only the boundary regions. This dimensional transition restores accuracy for critical areas while maintaining the speed benefits of low-resolution preliminary processing.
Solution Approach 2:
The low-resolution segmentation serves as a preliminary action that prepares boundary region identification without being the final processing step. The results of this preliminary low-resolution processing guide subsequent high-resolution processing of boundary areas, ensuring that the speed advantage of low-resolution processing does not compromise final extraction accuracy.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An image processing apparatus for extracting an area of a detection target from an image, the apparatus including: an image input section configured to acquire an image; an image generation section configured to generate a plurality of images with different resolutions from the image; and a segmentation section configured to perform segmentation using the plurality of images with the different resolutions, the segmentation section segmenting an image with a low resolution and then segmenting an image with a high resolution, using, as a processing target area, an area in the image with the high resolution corresponding to an area near a boundary resulting from processing of the segmentation of the image with the low resolution. This enables segmentation that is quicker and more accurate than in conventional techniques.