Electronic Document Privatization via Pixel Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for masking private information in electronic documents, such as emails, either distort the appearance of the documents or fail to effectively remove sensitive information, posing privacy issues during mass email campaigns and document storage.
Innovation Solution
A method that converts electronic documents into images, identifies and blurs private pixels, and generates a representative image by combining non-private portions with obfuscated imagery, ensuring private information is not visible while maintaining the document's appearance through a process involving image averaging and pixel manipulation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If private information is masked by blacking or whiting out, then private information is hidden, but the look and feel of the email presentation is degraded
Solution Approach 1:
The patent segments the image processing into distinct phases: identifying private pixels, creating a blurred version, generating a difference array, and selectively applying blur only to private regions. This segmentation allows precise control over which areas are obscured, protecting privacy while preserving the overall document structure and non-private content.
Solution Approach 2:
The patent applies different processing qualities to different regions of the image. Non-private areas maintain their original sharp quality, while only private pixel locations receive blur processing. This local differentiation ensures that privacy protection is applied precisely where needed without degrading the visual appearance of the entire document.
2Reliability
If current masking processes are used, then private information is obscured, but no visual information about the email is retained
Solution Approach 1:
The patent employs a dynamic, multi-step process that adapts to the specific characteristics of each image. By generating a blurred version, computing difference arrays to identify private regions, and selectively applying blur based on pixel differences, the system dynamically determines what to obscure rather than applying static masking rules, thereby preserving visual information where appropriate.
Solution Approach 2:
The patent introduces intermediate representations (blurred version, difference array) that serve as mediators between the original image and the final privatized output. These intermediaries allow the system to analyze and identify private information without permanently altering the original, enabling selective obscuration that maintains overall visual context.
3Stability of the object's composition
If a representative image is generated by averaging multiple images, then visual consistency is improved, but private information may persist across multiple images
Solution Approach 1:
The patent performs preliminary blurring of the averaged image before using it as a reference for identifying private pixels. By pre-blurring the representative image, the system ensures that even if private information appears consistently across multiple input images, it remains obscured in the reference, preventing leakage of private data while maintaining visual consistency for non-private areas.
Solution Approach 2:
The patent uses a feedback mechanism where the blurred representative image is compared against original images to identify private pixel locations. The difference array computation provides feedback that highlights regions where private information exists, allowing the system to iteratively refine its identification and ensure thorough privacy protection while maintaining visual consistency.
Data Source
AI summary
A set of electronic documents, such as emails, may include private information. A privatized image is generated to represent the set of electronic documents by identifying probable-pixel and proven-pixel locations within a subset of images. The subset of images is used to create blurred and obfuscated image having the private information distorted.