Merging Broken Table Lines via Cross Edge Overlap
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies face difficulties in accurately recognizing and merging broken table lines in handwritten tables due to irregularities such as smudged, broken, or faded segments, which can lead to errors in content recognition.
Innovation Solution
A method for image processing that identifies and merges broken line segments by determining overlap and applying a maximum width tolerance value, using cross edge pairs to assess the eligibility of segments for merging based on various tests and scoring systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If broken line segments are treated as separate entities, then each segment can be processed independently, but the table line recognition accuracy deteriorates due to fragmented information
Solution Approach 1:
The patent merges multiple broken line segments into a single continuous table line by detecting overlapping cross edges and calculating overlap ratios. When the overlap ratio exceeds a threshold, segments are combined into one unified line structure, restoring the continuity needed for accurate table recognition while maintaining processing efficiency through automated merging criteria
2Stability of the object's composition
If strict merging criteria are applied to combine broken segments, then line continuity is improved, but false merging of unrelated segments may occur
Solution Approach 1:
The patent replaces mechanical/visual inspection with mathematical cross edge detection and overlap ratio calculation. By computing the ratio of overlapping cross edges to total cross edges and comparing against a predefined threshold, the system objectively determines merging eligibility, reducing subjective judgment errors and false merges while ensuring reliable line continuity
3Measurement precision
If cross edge overlap detection is performed, then merging accuracy is improved, but processing time increases due to additional comparisons
Solution Approach 1:
The patent performs preliminary action by extracting and storing cross edge information during the initial line segmentation phase. By pre-processing the image to identify and store cross edge coordinates and characteristics before the merging decision phase, the system avoids time-consuming real-time calculations during the critical merging evaluation, thereby reducing overall processing time while maintaining high merging accuracy
Data Source
AI summary
A method for image processing includes obtaining a mask from an image; identifying, in the mask, a first stroke segment and a second stroke segment for possible merging; determining that the first stroke segment comprises a first plurality of cross edges that overlaps a second plurality of cross edges in the second stroke segment; identifying a plurality of cross edge pairs from the first plurality of cross edges and the second plurality of cross edges; determining a count of the plurality of cross edge pairs that satisfy a maximum width tolerance value of the first stroke segment; and merging the first stroke segment and the second stroke segment based on the count.