Document Image Entity Extraction Using Positional Reference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for extracting a character string corresponding to a predetermined item from a document image are inefficient when the character string is not present in the first document image, as they lack a method to accurately locate and extract the string from images with similar formats.
Innovation Solution
An image processing apparatus that performs character recognition on a document image, extracts character strings using entity extraction based on the recognition results, and if the string is not found, uses positional information from a second document image with the same format to locate and extract the string.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If entity extraction is performed only on the first document image, then the extraction process is simple, but the character string corresponding to the predetermined item cannot be extracted when it is not present in the first image
Solution Approach 1:
The system performs entity extraction on a second document image with the same format in advance, storing the extracted character string and its positional information. When processing the first document image, if the character string is not found, the pre-extracted information from the second image is utilized, avoiding the need for complex real-time search and extraction procedures.
Solution Approach 2:
The system creates a copy of the extracted character string and positional information from the second document image. This copied information is then used to supplement or replace extraction results from the first document image, ensuring accurate extraction without reprocessing the entire second image.
2Reliability
If the system searches for character strings in multiple document images, then extraction accuracy improves, but processing time increases
Solution Approach 1:
Entity extraction on the second document image is performed in advance and results are stored for future use. This preliminary action eliminates the need for repeated extraction operations on the same image, significantly reducing processing time while maintaining high extraction accuracy across multiple documents.
Solution Approach 2:
The system serves itself by storing extracted information from the second document image in a reusable format. When processing subsequent first document images, the system automatically utilizes this pre-stored information without requiring manual intervention or repeated extraction processes, thereby reducing processing time.
3Reliability
If the system uses positional information from a second document image, then extraction reliability improves when the string is missing from the first image, but the system complexity increases
Solution Approach 1:
The system copies positional information (coordinates, area data) from the second document image to locate and extract the character string when it is missing from the first image. This copying approach maintains extraction reliability without requiring complex image analysis or pattern recognition systems.
Solution Approach 2:
Positional information from the second document image serves as an intermediary that bridges the gap when extraction fails on the first image. This intermediary data structure contains pre-processed location and format information that simplifies the extraction process rather than complicating it.
Data Source
AI summary
When a character string corresponding to a predetermined item is not extracted in a first document image as a processing target by entity extraction processing, the character string corresponding to the predetermined item in the first document image is acquired based on positional information about an area where the character string corresponding to the predetermined item is previously extracted in a second document image having the same format as that of the first document image.