Character Direction Identification Using Language Consistency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for automatically determining the direction of characters in document images often misjudge the orientation due to disregarding language consistency, leading to decreased accuracy in optical character recognition processing.
Innovation Solution
A method that identifies the language group of characters in an image block by calculating the proportion of sub-image blocks belonging to specific language groups and adjusts the correctness measures for sub-image blocks not belonging to the determined language group, thereby improving the accuracy of direction determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If optical character recognition processing is performed in multiple directions to determine character orientation, then the automation of direction judgment is improved, but the accuracy of direction identification deteriorates due to language inconsistency being disregarded
Solution Approach 1:
The patent applies preliminary action by determining the language group of characters before performing optical character recognition in multiple directions. The system calculates the proportion of sub-image blocks belonging to specific language groups, identifies the dominant language, and then uses this language information to adjust correctness measures during direction determination. This preliminary language identification step ensures that the subsequent automated direction judgment process maintains high accuracy by considering language consistency across different orientations.
2Adaptability or versatility
If character recognition is performed across all possible orientations to find the best match, then adaptability to different document placements is improved, but the reliability of direction determination deteriorates due to false recognition in incorrect orientations
Solution Approach 1:
The patent applies local quality by adjusting the correctness measures of recognized characters based on their language group membership relative to the dominant language identified in the image block. Specifically, when a recognized character in a certain orientation does not belong to the dominant language group, its correctness measure is reduced. This localized adjustment of quality metrics for specific characters allows the system to maintain high adaptability to various document placements while improving reliability by penalizing false recognitions that result in incorrect language identification.
3Adaptability or versatility
If the system considers multiple languages and orientations comprehensively, then the versatility of the recognition system is improved, but the complexity of the processing increases
Solution Approach 1:
The patent applies segmentation by dividing the image block into multiple sub-image blocks and determining the language group for each sub-block. By calculating the proportion of sub-image blocks belonging to specific language groups, the system identifies the dominant language without needing to perform comprehensive recognition of all possible language-orientation combinations simultaneously. This segmented approach reduces processing complexity while maintaining versatility across multiple languages and orientations.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention discloses a method of and a device for identifying the direction of characters in an image block. The method includes: performing optical character recognition processing on the image block by assuming various directions as assumed character directions, respectively, to obtain sub image blocks, recognized characters corresponding to the sub image blocks and correctness measures thereof in each of the assumed character directions; determining a language group which the characters in the image block belong to; adjusting a correctness measure corresponding to a sub image block which corresponds to a recognized character not belonging to the determined language group in each of the assumed character directions; calculating an accumulative correctness measure in each of the assumed character directions based on the adjusted correctness measure; and identifying the direction of the characters in the image block according to the accumulative correctness measures.