Video Frame Face Replacement for PII Anonymization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video capture and storage systems fail to effectively anonymize personally identifiable information (PII) in real-time image data, leading to privacy concerns and compliance issues.
Innovation Solution
A method using generative adversarial networks (GANs) to generate and replace facial images in video frames, storing key-value pairings of vector representations to obfuscate recognizable faces, ensuring privacy and compliance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If facial images are captured and stored in video streams for security identification, then individual identification capability is improved, but privacy protection deteriorates due to storage of personally identifiable information
Solution Approach 1:
The patent extracts and removes personally identifiable facial information from video streams by detecting facial regions and replacing them with synthetic faces generated by GANs, thereby retaining video content utility while eliminating privacy risks associated with storing recognizable facial data
Solution Approach 2:
The patent creates synthetic face copies using generative adversarial networks that replicate the visual appearance of faces without containing the original personally identifiable information, allowing video streams to maintain realistic appearance while protecting individual privacy
2Object-affected harmful factors
If traditional face blurring or pixelation methods are used to protect privacy, then privacy protection is improved, but video quality and realism deteriorate
Solution Approach 1:
Instead of degrading the original face image through blurring or pixelation, the patent generates high-quality synthetic face copies using GANs that maintain visual realism and video quality while completely removing the original personally identifiable facial features
Solution Approach 2:
The patent transforms the facial data representation by converting real facial images into synthetic representations through GAN generation, changing the fundamental parameters of the face data while maintaining visual plausibility and video stream quality
3Object-affected harmful factors
If real-time facial obfuscation is implemented, then privacy protection is improved, but processing time and computational resources increase
Solution Approach 1:
The patent performs facial detection and GAN-based face generation in advance during video encoding or preprocessing stages, so that when video frames are displayed or stored, the facial obfuscation is already complete, reducing real-time processing delays
Solution Approach 2:
The system uses pre-trained GAN models that can autonomously generate synthetic faces without requiring manual intervention or complex real-time computation, enabling efficient automated privacy protection with reduced processing overhead
Data Source
AI summary
Obfuscating image data by receiving a plurality of video image frame data, detecting an object within a frame, generating a key associated with the object, matching the key to a key/value pair in a key store, generating a revised frame by replacing the object with the value, and providing the revised frame.