Edge Detection Algorithm Using Gradient Mini-Vector Thinning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current edge-detection algorithms in image processing systems produce unsatisfactory results when reducing edge zones to lines, as they fail to accurately define edges in images with less distinct features, leading to thick and poorly defined edge lines.
Innovation Solution
An edge-tracing process that analyzes pixel gradient magnitudes and directional orientations to identify mini-vectors, which are then refined to reduce edge line thickness to a single pixel, improving edge detection by locating the 'crest' line of high-gradient features in gradient sub-images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current edge-detection algorithms are used to detect edges in images, then edge zones can be identified, but the edge lines become thick and poorly defined
Solution Approach 1:
The patent segments the edge detection process into multiple stages: gradient calculation, mini-vector identification, and iterative thinning. By dividing the edge zone into mini-vectors (groups of connected high-gradient pixels) and progressively reducing them, the algorithm transforms thick edge zones into thin, well-defined edge lines while maintaining detection accuracy
Solution Approach 2:
The patent introduces directional orientation factors alongside gradient magnitude, adding a dimensional aspect to edge detection. By analyzing both the strength and direction of gradient changes, the algorithm can more precisely locate edge positions and reduce edge line thickness compared to magnitude-only approaches
2Productivity
If gradient thresholding is applied to reduce edge zones to lines, then edge detection speed improves, but edges with less distinct features are not accurately detected
Solution Approach 1:
The patent performs preliminary gradient calculation and directional analysis for all pixels before applying thresholding. By pre-computing gradient magnitudes and orientations, the algorithm prepares data structures that enable rapid subsequent processing through mini-vector identification and threshold-based filtering, achieving both speed and accuracy
Solution Approach 2:
The patent introduces mini-vectors as an intermediary structure between raw gradient data and final edge lines. Mini-vectors group connected high-gradient pixels with consistent directional orientation, serving as intermediate representations that facilitate both accurate edge localization and efficient processing through iterative thinning operations
3Reliability
If multiple pixels are used to represent an edge zone, then edge detection is more robust to noise, but the edge definition becomes less precise
Solution Approach 1:
The patent implements a dynamic, iterative thinning process that progressively reduces mini-vectors from multiple pixels to single-pixel edge lines. In each iteration, the algorithm identifies and removes pixels that do not satisfy edge criteria, dynamically adapting the edge representation until a precise single-pixel line remains, thereby resolving the trade-off between robustness and precision
Data Source
AI summary
The invention relates to a system and method of identifying an edge of an object in an image. The method comprises: identifying a segment of the image and a central pixel in the segment; identifying a set of mini-vectors as possible edges for the central pixel, utilizing a calculated gradient figure for the center pixel; and producing a reduced set of mini-vectors from the set of mini-vectors by removing mini-vectors that do not generate a continuous edge with an adjacent edge in an adjacent segment in the image.