OCR Reliability Calculation for Document Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing optical character recognition (OCR) technologies do not effectively evaluate the precision of recognition results, lacking a method to assess the reliability of the recognition process, which hinders accurate verification and potential double-checking of results.
Innovation Solution
An image-processing device and method that calculates the reliability of OCR results based on descriptive feature amounts of character strings within document images and outputs the results in a display mode reflecting this reliability, allowing for precise evaluation and potential double-checking.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If optical character recognition is applied to evaluate document content, then content evaluation capability is improved, but reliability assessment of recognition results deteriorates
Solution Approach 1:
The system introduces feedback mechanisms by calculating reliability scores based on descriptive feature amounts and using this information to adjust the verification process. High-reliability results skip manual verification while low-reliability results undergo double-checking, creating a closed-loop system that continuously improves recognition accuracy through selective feedback.
Solution Approach 2:
The system changes parameters by introducing reliability scores and descriptive feature amounts as new evaluation dimensions. Instead of treating all recognition results uniformly, the system varies the verification intensity based on calculated reliability parameters, transforming a single-parameter evaluation system into a multi-parameter assessment framework.
2Measurement precision
If manual verification of OCR results is performed to improve accuracy, then recognition precision is improved, but processing time increases
Solution Approach 1:
The system applies local quality by differentiating verification intensity based on local reliability characteristics. Instead of uniformly verifying all results, the system identifies specific low-reliability regions (individual character strings or fields) that require manual inspection, while leaving high-reliability regions unverified, thus optimizing the balance between precision and time consumption.
Solution Approach 2:
The system implements partial action by performing manual verification only on a subset of recognition results that fall below the reliability threshold. This selective approach avoids the excessive action of verifying every single result, reducing overall processing time while maintaining sufficient recognition precision through targeted verification of problematic cases.
3Measurement precision
If reliability calculation based on descriptive feature amounts is implemented, then evaluation accuracy is improved, but system complexity increases
Solution Approach 1:
The system segments the reliability assessment into distinct components: descriptive feature amount extraction, reliability score calculation, and verification decision-making. This segmentation allows each component to be independently optimized and managed, reducing overall system complexity while maintaining high evaluation accuracy through modular architecture.
Data Source
AI summary
An image-processing device includes: a reliability calculation unit configured to calculate reliability of a character recognition result for a document image which is a character recognition target on the basis of a descriptive feature amount of a character string of a specific item included in the document image; and an image output unit configured to output an image of the character recognition result indicating the character string of the specific item in a display mode in accordance with the reliability.