Character Recognition Device Likelihood Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing character recognition systems face challenges in accurately identifying characters in images, often requiring manual reattempts and corrections, which can be burdensome for users, especially when characters are hidden or mistakenly detected.
Innovation Solution
A character recognition device that processes input images by detecting candidate character areas, allowing users to correct likelihoods through simple point designation, thereby reattempting recognition with reduced burden and minimizing mistakes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual operations are required to reattempt recognition of failed characters, then recognition accuracy can be improved, but user burden increases
Solution Approach 1:
The system automatically performs reattempt recognition by utilizing the corrected candidate character areas and adjusted likelihood values without requiring manual user intervention. The computer service itself corrects its own recognition failures by processing the modified candidate areas again, thereby reducing user burden while maintaining improved recognition accuracy.
2Reliability
If multiple candidate character areas are processed, then recognition completeness improves, but processing complexity increases
Solution Approach 1:
The system changes the likelihood parameter values of candidate character areas based on correction information. By adjusting these probability parameters and comparing against a threshold, the system efficiently processes multiple candidates without complex algorithms, maintaining recognition completeness while controlling processing complexity through simple parameter adjustment and threshold comparison.
Data Source
AI summary
A CPU in a character recognition device detects at least one candidate character area having a likelihood higher than zero in the input image, the likelihood being a likelihood of including a character. The CPU determines that a candidate character area having a likelihood higher than a threshold as a character area. The CPU then superimposes the character area over the input image, and displays the resultant image. The CPU acquires a user input designating a first point on the input image. The CPU raises the likelihood of the candidate character area that is included in a first correction area including the first point. The CPU determines a candidate character area having a likelihood higher than a threshold as a character area again. The CPU then superimposes the character area over the input image and displays the resultant image again. The CPU recognizes the characters included in the character areas.