Character Image Transformation for OCR Accuracy

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current optical character recognition (OCR) and intelligent character recognition (ICR) systems face challenges in accurately distinguishing between similar characters and fonts, leading to lower than desired recognition results, especially in real-world documents with varying font sizes, styles, and placements.

Innovation Solution

A system and method that transforms character images from one representation to another, allowing for font-to-font or handwriting-to-font conversion, identifies and separates background and content data, and translates character images to correct locations, enabling more accurate recognition by converting handwritten or typed fonts into uniform types, thereby improving legibility and recognition accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional OCR systems use fixed font types (OCR-A, OCR-B) for character recognition, then the recognition process is standardized and machine-readable, but the systems struggle to distinguish between similar characters and fonts in real-world documents with varied typography

Engineering Contradiction:
Improvecharacter recognition accuracyVSAvoidfont variation handling
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The system transforms character images by modifying their visual parameters (font type, size, spacing) to convert diverse handwriting and typed fonts into a uniform representation that OCR can accurately recognize, thereby resolving the contradiction between recognition precision and adaptability to font variations

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent introduces an intermediary transformation step between the original character image and OCR processing. This intermediary system analyzes the input character and converts it to a standardized font representation, acting as a mediator that enables accurate OCR recognition while handling diverse input fonts

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If OCR systems rely on specific font characteristics for recognition, then recognition speed is maintained, but recognition accuracy decreases when encountering unexpected font styles, sizes, or placements

Engineering Contradiction:
Improverecognition consistencyVSAvoidfont analysis requirements
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The system performs preliminary font analysis and transformation before the main OCR recognition process. By pre-converting character images to a uniform font representation and analyzing surrounding content for context, the system ensures consistent recognition results while managing complexity through staged processing

Inventive Principle:
Principle #10Preliminary action

3Measurement precision

If the system transforms all character images to uniform fonts, then recognition accuracy improves, but processing time and computational resources increase

Engineering Contradiction:
Improvecharacter identification accuracyVSAvoidprocessing efficiency
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The system applies font transformation selectively rather than uniformly to all characters. By analyzing the input and applying transformations only where necessary to achieve accurate recognition, the system maintains high precision while optimizing processing efficiency through targeted rather than exhaustive transformation

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentUS10586133B2System and method for processing character images and transforming font within a document
Publication Date: 2020.03.10 NLP LOGIX LLC
  • US10586133B2 patent drawing
  • US10586133B2 patent drawing
  • US10586133B2 patent drawing

AI summary

The present disclosure relates to a system and method to transform character images from one representation to another representation. According to some embodiments of the present disclosure, a form may be processed to separate background data from content data, wherein character images from one or both the background data and the content data may be transformed. In some aspects, one or both handwritten font and type font may be processed in the character images, wherein the original fonts may be transformed into a uniform type font. In some embodiments, the character images may be translated to their correct state, wherein the translation may occur before or after the transformation. In some implementations, the translation and font transformation may allow for more efficient and effective character recognition.