Document Edge Segmentation Using Multi-Method Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current segmentation techniques for digital images captured using cameras are inadequate in separating document images from surrounding images, especially under less than optimal conditions such as shadows, poor lighting, and image distortions, leading to poor segmentation rates.
Innovation Solution
A method and system utilizing multiple edge detection techniques to determine the edges of a document image, grouping potential edges into subsets, and selecting the most accurate edge sets based on predefined criteria such as trapezoidality, area, contrast, texture, and line extension or shrinkage, to effectively segment the document from the surrounding image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple edge detection techniques are used to improve segmentation accuracy, then segmentation precision improves, but system complexity increases
Solution Approach 1:
The patent divides the edge detection process into multiple independent techniques (Canny edge detection, Sobel filter, gradient-based detection) that operate separately on the digital image. Each technique generates its own edge map, which is then processed independently through grouping and selection stages. This segmentation of the detection process allows for improved accuracy through multiple perspectives while maintaining modular complexity that can be managed through systematic processing.
Solution Approach 2:
The patent combines the results from multiple edge detection techniques by grouping edges from different methods into candidate edge sets. These separate detection results are merged through a selection process that evaluates multiple criteria (trapezoidality, area, contrast, texture) to identify the most accurate document edges. This merging approach leverages the strengths of different detection methods while filtering out false positives through comprehensive evaluation.
2Reliability
If multiple edge detection techniques and selection criteria are applied, then segmentation reliability improves, but processing time increases
Solution Approach 1:
The patent performs preliminary grouping of edges from multiple detection techniques into organized candidate sets before the final selection process. By pre-organizing the edge data according to spatial relationships and detection method characteristics, the system reduces the computational burden of the subsequent selection stage. This preliminary structuring ensures reliable segmentation while minimizing processing time through efficient data organization.
Solution Approach 2:
The patent evaluates multiple edge sets against several criteria (trapezoidality, area, contrast, texture) but does not require all criteria to be perfectly satisfied. Instead, the system selects the edge set that best satisfies the combined criteria, accepting that no single set may perfectly meet all requirements. This partial satisfaction approach maintains high reliability while avoiding excessive processing time that would result from requiring perfect fulfillment of all criteria.
3Measurement precision
If edge grouping and selection based on multiple criteria is performed, then segmentation precision improves, but computational complexity increases
Solution Approach 1:
The patent applies different evaluation criteria to different aspects of edge sets: trapezoidality assesses geometric shape quality, area evaluates spatial extent, contrast measures intensity differentiation, and texture analyzes local patterns. Each criterion focuses on a specific local property of the edges rather than requiring a single complex global evaluation. This localized assessment approach improves precision through comprehensive evaluation while managing computational complexity by breaking down the assessment into independent, specialized metrics.
Data Source
AI summary
The present application provides an improved segmentation method and system for processing digital images that include an imaged document and surrounding image. A plurality of edge detection techniques are used to determine the edges of the imaged document and then segment the imaged document from the surrounding image.