Font Attribute Detection via Dense Image Normalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing OCR technologies face challenges in accurately detecting font styles such as bold, italic, and underline due to variations in font types and languages, leading to mischaracterizations, especially when thinner lettering is mistaken for emboldened text.
Innovation Solution
The method involves generating a dense image document by normalizing text size, removing blank spaces, and altering character positions to analyze font attributes using a semantic segmentation model, which aggregates characteristics and annotates words with accurate font attributes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional OCR methods are used to detect font styles, then the process is simple and fast, but the accuracy is low due to mischaracterizations from font variations
Solution Approach 1:
The patent segments the font detection task into multiple independent analysis dimensions: stroke width analysis, character shape analysis, spacing analysis, and position analysis. Each dimension is processed separately to extract specific features, which are then combined to determine the final font style. This segmentation allows comprehensive analysis without requiring a single complex detection model.
Solution Approach 2:
The patent transforms the 2D image data into multiple analysis dimensions by examining stroke width variations, character shape properties, spacing relationships, and positional information. This multi-dimensional approach converts a single complex classification problem into several simpler analytical dimensions that can be processed independently and then integrated.
2Adaptability or versatility
If text size and spacing variations are present in the document, then the document represents real-world diversity, but font style detection accuracy deteriorates due to mischaracterizations
Solution Approach 1:
The patent applies local quality analysis by examining specific local features of characters rather than treating the entire character uniformly. It analyzes stroke width at different locations within characters, examines local shape properties of individual strokes, and evaluates spacing relationships between adjacent characters. This local-focused approach allows accurate font style detection even when overall text size and spacing vary.
Solution Approach 2:
The patent changes the parameters being analyzed from absolute measurements to relative and normalized parameters. Instead of measuring absolute stroke widths or spacing distances, it analyzes stroke width ratios, relative spacing relationships, and normalized character proportions. This parameter transformation makes the detection invariant to text size and spacing variations while maintaining sensitivity to font style differences.
3Reliability
If thin lettering is present in the document, then the document maintains readability, but font style detection fails by mistaking thin lettering for emboldened text
Solution Approach 1:
The patent performs preliminary analysis of stroke width characteristics and character shape properties before making font style determinations. By预先 examining the distribution and consistency of stroke widths across multiple characters, and analyzing the structural properties of character shapes, it establishes a baseline understanding that prevents misclassification of thin lettering as bold text. This preliminary characterization is crucial for reliable font style detection.
Solution Approach 2:
The patent incorporates feedback mechanisms by using the results from multiple analysis dimensions (stroke width, shape, spacing, position) to validate and refine font style determinations. The system cross-checks findings across different analytical dimensions, and uses inconsistencies as feedback to re-evaluate potential misclassifications. This feedback loop significantly improves reliability in distinguishing thin lettering from actual bold text.
Data Source
AI summary
Described are techniques for font attribute detection. The techniques include receiving a document having different font attributes amongst a plurality of words respectively comprised of at least one character. The techniques further include generating a dense image document from the document by setting the plurality of words to a predefined size, removing blank spaces from the document, and altering an order of characters relative to the document. The techniques further include determining characteristics of the characters in the dense image document and aggregating the characteristics for at least one word. The techniques further include annotating the at least one word with a font attribute based on the aggregated characteristics.