Text Character Identification System Using Adaptive Filtering Rules
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional text extraction methods from bitmap images face reduced accuracy due to variations in text size, color, background images, and oblique text, especially when text color is similar to the background, and resolution-dependent parameters can lead to incorrect results.
Innovation Solution
A system and method for text character identification in binary images that acquires connected components, determines property limits, generates filtering rules based on these limits, and classifies components as text characters, using a connected component acquisition unit, limit generation unit, filtering rule generation unit, and text determination unit to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional text extraction algorithms (bottom-up or top-down) are used, then text can be extracted from bitmap images, but accuracy is reduced due to variations in text size, color, background images, and oblique images
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting extraction parameters based on image characteristics. The system analyzes image resolution, text density, and color contrast to automatically adapt extraction parameters, thereby maintaining high accuracy across varying text sizes, colors, and background conditions without requiring manual parameter tuning for each scenario.
Solution Approach 2:
The patent implements dynamics by creating an adaptive extraction system that dynamically adjusts its behavior based on real-time image analysis. The system continuously evaluates image properties and modifies extraction parameters during processing, enabling it to handle diverse text variations (size, color, orientation, background) while maintaining consistent accuracy across different document types.
2Measurement precision
If predetermined parameters are adjusted according to resolution variation, then extraction accuracy is improved, but lower resolution images may not contain smaller text size and thus adjustment may cause incorrect results
Solution Approach 1:
The patent applies feedback by implementing a validation mechanism that checks extraction results against image characteristics. The system analyzes the extracted text and compares it with expected patterns based on image resolution and other properties, then adjusts parameters accordingly. This feedback loop ensures that parameter adjustments are validated and corrected if they lead to incorrect results, maintaining both accuracy and reliability across different resolutions.
Solution Approach 2:
The patent implements preliminary action by performing comprehensive image analysis before text extraction. The system pre-processes the image to determine actual text characteristics (size, density, contrast) and uses this information to set appropriate extraction parameters before the actual extraction occurs. This preliminary characterization prevents incorrect parameter adjustments that would result from relying solely on resolution-based assumptions.
3Measurement precision
If multiple conversion methods are performed to maximize contrast between text and background, then text extraction is attempted, but text near background color cannot be correctly extracted
Solution Approach 1:
The patent applies parameter changes by using advanced color space transformations and adaptive thresholding techniques. Instead of relying on simple contrast maximization through multiple conversions, the system transforms images into multiple color spaces (RGB, HSV, LAB) and dynamically selects the most effective transformation parameters based on local image characteristics. This allows accurate extraction of text with colors similar to the background by adapting parameters to local color distributions rather than applying fixed conversion sequences.
Data Source
AI summary
A method for text character identification. The method acquires multiple connected components (CCs) in a binary image, and each CC has a pattern property value. The method determines at least one property limit based on the pattern property values, generates a filtering rule according to the property limit, and determines whether each of the CCs is a text character according to the filtering rule.