Image Segmentation Using Curvature-Based Watershed Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image segmentation techniques often suffer from over-segmentation, leading to inefficiencies in processing and analysis, as they fail to effectively differentiate between object and background areas due to arbitrary pixel groupings and noise.
Innovation Solution
A computer-implemented image segmentation method that evaluates pixel curvature, defines a topographic relief, calculates a watershed to delimit catchment basins, and partitions pixels based on this watershed, thereby achieving a geometrically natural segmentation by discretizing curvature values into a finite number of discrete values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional image segmentation techniques are used to partition pixels into groups, then the image can be segmented into multiple regions, but over-segmentation occurs leading to too many groups and reduced processing efficiency
Solution Approach 1:
The patent transforms the segmentation approach by changing from intensity-based parameters to curvature-based parameters. By computing curvature values for each pixel and using these as the basis for segmentation, the method achieves more meaningful region boundaries that reduce over-segmentation while maintaining processing efficiency through the mathematical properties of curvature.
Solution Approach 2:
The patent replaces traditional mechanical grouping methods (arbitrary pixel clustering based on intensity thresholds) with a geometric approach based on curvature analysis. This substitution allows regions to be defined by intrinsic geometric properties rather than arbitrary parameters, naturally reducing the number of segments while improving accuracy.
2Ease of operation
If arbitrary pixel groupings are used for segmentation, then processing can be simplified, but the segmentation becomes faulty and does not accurately represent object geometry
Solution Approach 1:
The patent changes the fundamental parameter used for segmentation from pixel intensity to pixel curvature. This parameter transformation enables the segmentation to respect geometric boundaries while maintaining computational simplicity through efficient curvature calculation algorithms.
Solution Approach 2:
The patent introduces a new dimension of analysis by computing curvature values that capture the geometric structure of the image. This additional geometric dimension allows the segmentation to naturally align with object boundaries without requiring complex processing steps.
3Measurement precision
If continuous curvature values are used for segmentation, then geometric precision is maintained, but the complexity of processing increases due to the large number of possible curvature values
Solution Approach 1:
The patent applies segmentation to the curvature values themselves by discretizing the continuous curvature domain into a finite set of representative values. This allows the method to maintain geometric precision while reducing processing complexity through the use of discrete curvature levels in the watershed transformation.
Solution Approach 2:
The patent transforms the continuous curvature parameter into a discrete parameter space suitable for digital image processing. By mapping continuous curvature values to a finite set of discrete levels, the method preserves geometric information while enabling efficient computational processing.
Data Source
AI summary
The invention relates to a computer-implemented segmentation method of an image comprising a segmentation by watershed applied to an evaluation of a curvature at pixels of the image.