Image Classification Masking for Confidential Data Protection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In image classification processes, incorrect classification of documents containing confidential information leads to increased user effort and time in investigating the cause, as existing methods often degrade image quality when masking sensitive data, potentially losing essential information.

Innovation Solution

An image processing system that includes an image scanning apparatus and a terminal apparatus, which perform classification, allow users to correct classification errors, and implement a masking process to generate a secure image by masking regions outside the specified classification rules, ensuring confidential information is hidden while maintaining necessary information for investigation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Object-affected harmful factors

If masking processing is performed on images containing confidential information, then security is improved, but image quality deteriorates and essential information may be lost

Engineering Contradiction:
Improveconfidential information protectionVSAvoidessential information loss
Core Design Contradiction:
Object-affected harmful factorsVSLoss of information

Solution Approach 1:

The patent divides the image into multiple regions based on classification rules. Only specific regions containing confidential information are masked, while other regions preserving essential information for cause investigation remain unmasked. This selective segmentation approach resolves the contradiction by applying masking only where necessary for security while maintaining information availability where needed for investigation.

Inventive Principle:
Principle #1Segmentation

2Object-affected harmful factors

If user performs manual masking processing to protect confidential information, then security is improved, but user time and effort increase

Engineering Contradiction:
Improveconfidential information protectionVSAvoiduser time for cause investigation
Core Design Contradiction:
Object-affected harmful factorsVSLoss of time

Solution Approach 1:

The system automatically performs masking processing based on pre-stored classification rules without requiring manual user intervention. The classification result determination unit and masking processing unit work autonomously to identify and mask confidential information regions, eliminating the need for users to manually perform time-consuming masking operations while still achieving secure protection.

Inventive Principle:
Principle #25Self-service

3Productivity

If classification processing is performed on images with confidential information, then organization is improved, but incorrect classification occurs requiring additional user investigation

Engineering Contradiction:
Improveimage classification efficiencyVSAvoidclassification accuracy
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent incorporates a feedback mechanism where the determination unit receives classification results and feeds them back to the masking processing unit. This feedback loop allows the system to adjust masking operations based on classification outcomes, enabling correction of incorrect classifications while maintaining overall processing efficiency. The feedback mechanism ensures that classification errors can be identified and addressed without compromising productivity.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP3147810B1Image processing apparatus and program
Publication Date: 2019.08.28 FUJIFILM BUSINESS INNOVATION CORP
  • EP3147810B1 patent drawingFigure 1
  • EP3147810B1 patent drawingFigure 2
  • EP3147810B1 patent drawingFigure 3

AI summary

An image processing apparatus includes a classifying unit and a replacing unit. The classifying unit classifies an input image into a category among plural categories. The replacing unit replaces a portion of the input image with a first image in a case where the category into which the input image is classified by the classifying unit is corrected by a user to a correct category among the plural categories.