Font Identification via Relative Line Width Fingerprints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for font identification in documents are time-consuming and unreliable, especially with distorted reproductions, as they require analyzing individual glyph properties and comparing them, which is cumbersome and prone to errors.
Innovation Solution
A computer-implemented method that generates fingerprints based on relative line widths of text set in a proportional font, allowing for automated comparison between original and test text to determine font consistency, even with low-quality reproductions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If individual glyph properties are examined and compared to identify fonts, then font identification can be performed, but the process becomes time-consuming and requires extensive analyst study time
Solution Approach 1:
The patent segments the font identification process into distinct components: extracting glyph images from the document, generating fingerprints from these glyphs, and comparing fingerprints to identify the font. This segmentation automates the previously manual analyst process, reducing time consumption while maintaining identification accuracy through systematic processing of individual glyph characteristics
Solution Approach 2:
The patent replaces the mechanical/manual process of analyst study and comparison with an automated computer-based system that extracts glyph images, generates fingerprints algorithmically, and performs automated pattern matching. This substitution eliminates manual time investment while maintaining or improving identification reliability through consistent automated analysis
2Reliability
If individual glyph properties are examined to identify fonts, then font identification can be performed, but the process becomes complex and cumbersome
Solution Approach 1:
The patent extracts only the essential characteristics needed for font identification by generating fingerprints from glyph images. Instead of examining all individual glyph properties in detail, the system extracts key features that define font characteristics, simplifying the overall process while maintaining identification accuracy through focused analysis of discriminative features
Solution Approach 2:
The patent transforms the complex set of individual glyph properties into a simplified fingerprint representation. By changing the parameter space from multiple detailed glyph measurements to a condensed fingerprint format, the system reduces process complexity while preserving the essential information needed for accurate font identification through pattern matching
3Reliability
If glyphs in distorted document reproductions are analyzed, then font identification may be attempted, but the original glyphs are too distorted to provide reliable identification
Solution Approach 1:
The patent performs preliminary processing of the distorted glyph images by extracting and generating fingerprints before comparison. This preliminary action prepares the distorted data in a format that emphasizes invariant characteristics, allowing the system to identify fonts even when original glyphs are distorted by reproduction processes through pre-processing that highlights enduring font features
Solution Approach 2:
The patent converts the distortion present in reproduced documents into a benefit by using fingerprint generation that focuses on relative relationships between glyph features rather than absolute measurements. This approach transforms the harmful distortion into an opportunity to identify fonts based on proportional relationships that remain consistent even when overall glyph shapes are degraded by reproduction
Data Source
AI summary
A computer-implemented method of font identification includes receiving a first document, the first document including the first text set in a proportional font. Test text, corresponding to the first text of the first document, is received. The test text is set in a test font. A first fingerprint is generated, based on relative line widths of the first text of the first document. A second fingerprint is generated based on relative line widths of the test text, as set in the test font. The test font is then accepted as being consistent with a font of the first text, based on a predetermined strength of relationship between the first and second fingerprints.