Table Grid Detection and Separation in Scanned Documents
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Analyzing and recognizing the content of scanned documents, particularly tables within them, is challenging due to variations in table designs and issues like skewing or handwriting, which complicates the parsing of document content.
Innovation Solution
A technique for detecting and separating table grids from scanned document images involves extracting connected components, identifying table grid candidates, extracting inverse and regular cells, and separating noise, graphics, or handwriting from the table grid, using methods such as binary image generation, smearing operations, and validation tests to determine valid grid components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional table detection methods are used on scanned documents, then simple table structures can be detected, but documents with variations in table designs, skewing, or handwriting cannot be accurately processed
Solution Approach 1:
The patent segments the table detection process into multiple distinct phases: connected component extraction, inverse cell detection, grid line identification, and validation. Each phase handles specific aspects of table structures, enabling the system to process various table designs while maintaining accuracy through specialized processing for each segment.
Solution Approach 2:
The patent employs multiple validation tests that check different parameters of potential table grids including rectangularity degree, variance of run lengths, density thresholds, and grid line component ratios. By evaluating multiple parameters simultaneously, the system adapts to different table designs while maintaining precise detection through quantitative criteria.
2Measurement precision
If complex validation tests are applied to detect table grids accurately, then detection precision improves, but processing time and computational complexity increase
Solution Approach 1:
The patent performs preliminary processing by extracting connected components and identifying inverse cells before conducting grid line detection and validation. This preliminary segmentation reduces the search space for subsequent validation tests, allowing accurate detection while reducing computational overhead by pre-organizing the data structure.
Solution Approach 2:
The patent applies different validation criteria to different regions and components of the table structure. For example, inverse cells are validated using rectangularity and run length variance specific to their location, while grid lines are validated using density and component ratio tests appropriate for line structures. This localized validation approach maintains precision while reducing unnecessary computations.
3Productivity
If all connected components are processed to extract table grids, then complete table detection is achieved, but noise, graphics, and handwriting are also incorrectly identified
Solution Approach 1:
The patent extracts and removes inverse cells (white holes within connected components) before proceeding with grid line detection. This extraction step separates the table structure from surrounding noise, graphics, or handwriting that do not form inverse cells, thereby improving reliability by eliminating false positives while maintaining complete table detection through subsequent grid line validation.
Solution Approach 2:
The patent implements multiple validation tests that provide feedback on whether detected components are genuine table grids. The validation checks include rectangularity degree, variance of run lengths, density thresholds, and grid line component ratios. Components that fail these feedback-based validation tests are rejected, reducing false positives while maintaining complete detection of valid tables through iterative validation.
Data Source
AI summary
A technique is described for table grid detection and separation during the analysis and recognition of documents containing table contents. The technique includes the steps of table detection, grid separation, and table cell extraction. The technique is characterized by the steps of detecting the grid lines of a table using, for example, inverse cell detection, separating noise and touching text from the grid lines, and extracting the cell contents for OCR recognition.