Hand-Drawn Table Geometry Reconstruction via Line Clustering

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Image processing devices face difficulties in determining the geometry of hand-drawn tables due to non-straight boundaries, making it challenging to generate a high-level representation for electronic documents.

Innovation Solution

A method involving identifying geometric lines in an image using the Hough Transform, clustering them, selecting the strongest lines, calculating points, and reconstructing the table geometry to determine the layout, including the upper left corner, number of rows and columns, and cell positions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If hand-drawn lines are used to represent table boundaries, then the table can be freely drawn and edited, but the lines become non-straight and make it difficult to determine table geometry

Engineering Contradiction:

Improvefreedom to draw and edit tablesVSAvoidtable geometry determination accuracy

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent segments the hand-drawn lines into multiple geometric line segments using the Hough Transform. Each curved or irregular line is broken down into straight line segments that can be individually processed and analyzed, enabling precise geometric representation while preserving the original hand-drawn appearance.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent creates a geometric copy or representation of the hand-drawn lines. By applying the Hough Transform, the system generates idealized geometric line segments that replicate the essential structure of the original hand-drawn lines, allowing for accurate geometry determination while maintaining the flexibility of hand-drawn input.

Inventive Principle:

Principle #26Copying

2Reliability

If multiple geometric lines are identified in the image, then more potential table boundaries are captured, but it becomes difficult to distinguish actual table lines from false positives

Engineering Contradiction:

Improvedetection of potential table linesVSAvoidline selection and filtering process

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent employs a feedback mechanism where detected geometric lines are evaluated based on their spatial relationships and consistency with table structure. Lines that do not conform to expected table geometry patterns are filtered out, while valid lines are reinforced and selected, creating an iterative refinement process that reduces false positives.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent changes parameters such as line confidence thresholds, spatial tolerance values, and geometric consistency criteria to optimize the distinction between actual table lines and false positives. By adjusting these parameters, the system can adapt to different image qualities and table configurations, improving reliability while managing complexity.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20170193292A1Identifying the lines of a table

Publication Date: 2017.07.06 KONICA MINOLTA SYSTEMS LABORATORY INC

AI summary

A method for image processing, including: obtaining an image including a table; identifying a first plurality of geometric lines in the image; grouping the first plurality of geometric lines into a plurality of clusters; determining a plurality of hand-drawn lines in the image corresponding to the table from the plurality of clusters; calculating a plurality of points for the plurality of hand-drawn lines; and determining a geometry of the table based on the plurality of points.