Bitmapped Text Font Size Detection Without OCR
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining font size in electronic documents are inconsistent due to variations in character cases, ascenders, descenders, accents, and font types, and require optical character recognition (OCR) or prior knowledge of font type, leading to unreliable results.
Innovation Solution
A machine learning model trained on bitmapped text with known point sizes uses Gabor filters and deep convolutional neural networks to determine font size independently of font type, relying on feature extraction and scaling techniques to normalize text dimensions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a bounding box is placed around rendered text to determine font point size, then the measurement process is simple, but the determination is inaccurate due to variations in character cases, ascenders, descenders, and accents
Solution Approach 1:
The patent replaces the mechanical bounding box measurement approach with a machine learning-based optical recognition system. The ML model analyzes visual features of text directly from images, substituting the simple but inaccurate geometric bounding box method with a more complex but accurate perceptual system that understands text structure and semantics.
Solution Approach 2:
The patent transforms the measurement parameters from simple geometric bounding box dimensions to complex visual features extracted by the ML model, including text height, width, aspect ratio, and other morphological characteristics that better represent actual font size while accounting for character variations.
2Reliability
If optical character recognition (OCR) is used to determine font size, then character identification is possible, but the process becomes complex and requires prior knowledge of font type
Solution Approach 1:
The patent extracts only the necessary visual features for font size determination from the full OCR process. Instead of performing complete character recognition and font type identification, the system extracts relevant geometric and morphological features from text images and feeds them directly to the ML model, eliminating unnecessary processing steps.
Solution Approach 2:
The patent creates a simplified representation of text by extracting key visual features rather than performing full OCR. This feature extraction creates a copy of the essential information needed for font size determination without requiring complete character identification or font type knowledge.
3Measurement precision
If font type and variant information is used to determine font size, then some accuracy can be achieved, but the method fails when font type is unknown or varies across different text elements
Solution Approach 1:
The patent creates a universal ML-based measurement system that works across all font types, sizes, and styles without requiring font-specific calibration. The model is trained on diverse text images and learns generalizable features that apply universally, making the system adaptable to any font while maintaining measurement accuracy.
Solution Approach 2:
The patent performs preliminary training of the ML model on large datasets of text images with known font sizes before deployment. This preliminary action allows the system to learn the relationships between visual features and actual font sizes across various fonts, preparing it to handle unknown or varying font types without requiring prior font identification.
Data Source
AI summary
Method and apparatus for determining font point size in bitmapped text does not rely on accuracy of an optical character recognition (OCR) engine, or on generation of heuristics (e.g. assumption of certain amounts of different types of text, such as capital, lowercase, ascending, descending) to determine a likely font size. A deep learning model for determining text size is based on extraction of features from existing text to obtain a more general solution.