Image Data Masking via Classification-Based Category Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for masking processing in image data are unable to effectively handle character strings that differ by type, as seen in Japanese Patent Application Laid-Open No. 2012-234344, which limits the ability to perform masking on varied image data sets.
Innovation Solution
An information processing apparatus that includes a processor for identifying classifications of image data, searching for character strings within specific categories, and performing masking processing on those strings, allowing for different categories associated with various types of image data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a single masking processing technique is used for all image data, then the processing method is simple, but it cannot handle character strings that differ by type of image data
Solution Approach 1:
The patent segments the masking processing system into multiple classification units, each specialized for a specific type of image data (e.g., invoice processing unit, receipt processing unit). Each unit contains classification means tailored to its specific image type, enabling adaptable handling of different data formats while maintaining modular system architecture that prevents excessive complexity.
Solution Approach 2:
The patent implements a universal masking processing framework that can handle multiple types of image data through a common structure. The system uses a unified processing flow that routes different image types to appropriate classification units, allowing the system to be versatile across various document types without requiring completely separate processing paths for each type.
2Measurement precision
If masking processing is performed on all character strings uniformly, then the processing is straightforward, but sensitive information of different types cannot be accurately identified and concealed
Solution Approach 1:
The patent applies local quality by implementing type-specific classification rules and search criteria for each image data category. For example, invoice processing uses classification rules appropriate for financial documents, while receipt processing uses different rules suited for transaction records. This allows accurate identification of sensitive information specific to each document type while maintaining a standardized overall processing framework.
Solution Approach 2:
The patent introduces classification means as an intermediary layer between the image data input and the masking execution. This intermediary classifies the image data by type and selects appropriate search criteria, thereby improving the accuracy of sensitive information identification without requiring the entire system to be restructured for each document type, thus managing complexity effectively.
3Productivity
If recognition processing is performed on all image data without classification, then the processing flow is simple, but character strings corresponding to items that differ for each type cannot be searched
Solution Approach 1:
The patent performs classification of image data by type before executing the masking processing. This preliminary action allows the system to select appropriate search criteria and classification rules specific to each document type, thereby improving the efficiency of subsequent character string search and masking operations. The classification step is integrated into the processing flow to minimize additional complexity.
Data Source
AI summary
An information processing apparatus performs operations including, based on recognition processing for identifying a classification of image data, identifying a classification of the image data from among a plurality of classifications, searching for a character string falling into a category associated with the identified classification from among a plurality of character strings included in the image data, and performing masking processing on the searched-for character string in the image data, wherein each of the plurality of classifications are associated with different categories.