Document Logical Structure Recovery via Vertical Cut Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining the logical ordering of layout objects in document images often fail to accurately restore the reading order, especially when documents are scanned or optimized for printing, storage, or display, leading to incorrect or lost structural information.
Innovation Solution
A method that assigns a page as a block, identifies hypothetical cuts between layout objects, computes arrangement criteria, and modifies these cuts to determine the logical order by merging cuts that span multiple sub-blocks, favoring vertical cuts over horizontal ones to establish a correct reading order.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If documents are optimized for printing, storage, or display formats, then print/storage/display efficiency is improved, but logical document structure is lost or degraded
Solution Approach 1:
The patent segments the document into layout objects and identifies cuts between them to reconstruct the logical reading order. By dividing the document into discrete objects and analyzing the spatial relationships between them, the system can recover the logical structure even when the original high-level format is lost.
Solution Approach 2:
The patent transforms the two-dimensional spatial arrangement of layout objects into a one-dimensional logical reading order. By analyzing the positions and relationships of objects in the 2D page layout, the system reconstructs the 1D sequence in which content should be read, effectively converting spatial information into logical ordering information.
2Adaptability or versatility
If scanned hardcopy documents are converted to electronic form, then document accessibility is improved, but document structure is lost unless augmented with high-level description
Solution Approach 1:
The patent enables scanned documents to self-describe their logical structure by analyzing the geometric and spatial relationships of layout objects within the document itself. Instead of requiring external high-level descriptions, the system extracts structural information directly from the document's visual layout, allowing the document to serve its own structuring needs.
3Measurement precision
If known methods exploit geometric or typographical features to determine logical structure, then structure recovery is possible, but accuracy fails for fine-grain layout objects and varying orientations
Solution Approach 1:
The patent employs a dynamic approach by iteratively identifying and modifying cuts based on arrangement criteria. Rather than using a fixed rule set, the system adapts its analysis by repeatedly examining layout objects, identifying optimal cuts, and refining the reading order determination, which improves accuracy for complex layouts with fine-grain objects and varying orientations.
Data Source
AI summary
Methods are disclosed for recovering or determining logical structure of a document by assessing different combinations of vertical and horizontal cuts across a block of the document. The block is segmented using a scoring function that discards horizontal cuts in favor of vertical cuts shared among neighboring sub-blocks. The order in which the blocks and sub-blocks are segmented is then used to define the logical structure of the document, such as its reading order.