Reversible Face Redaction via Non-Uniform Block Scrambling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Vehicle passenger detection systems face challenges in implementing effective face redaction solutions that balance privacy protection with the need to identify passengers, particularly in real-world applications where privacy concerns hinder the full utilization of computer vision applications.
Innovation Solution
A method and system for reversible face redaction in vehicle passenger detection systems, involving image processing techniques that divide regions of interest into non-uniform blocks, scramble each block, and use coordinates to redact and reconstruct faces, ensuring privacy while maintaining image intelligibility and allowing for reversible processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If face redaction is applied to protect privacy, then privacy protection is improved, but the ability to identify passengers deteriorates
Solution Approach 1:
The face image is divided into multiple non-uniform blocks of different sizes. This segmentation allows selective processing where different blocks can be redacted with different levels of intensity, preserving some structural information while protecting privacy. The non-uniform block sizes create a more complex redaction pattern that maintains intelligibility better than uniform blocking.
Solution Approach 2:
Different regions of the face are treated differently through non-uniform block division. Critical privacy-sensitive areas (like eyes, mouth) are covered with larger blocks, while less sensitive areas use smaller blocks. This local differentiation maintains overall face structure intelligibility while providing targeted privacy protection.
2Reliability
If original images are retained for law enforcement purposes, then legal compliance is improved, but privacy violations increase
Solution Approach 1:
A reversible redaction mechanism serves as an intermediary between privacy protection and law enforcement needs. The redaction process embeds reversible information that allows authorized entities to restore original images when legally required, while maintaining privacy protection during normal operation. This intermediary solution satisfies both conflicting requirements.
Solution Approach 2:
The redaction is applied in advance to all images before storage or transmission, preparing them for privacy protection. However, the redaction process preserves the capability for reversible restoration, so when legal compliance is needed, the original information can be recovered without needing to retain unredacted copies.
3Ease of manufacture
If uniform block redaction is used, then implementation simplicity is improved, but redaction effectiveness deteriorates
Solution Approach 1:
The patent employs non-uniform block sizes instead of uniform blocks for redaction. This asymmetry in block dimensions creates more varied coverage patterns that are more effective at protecting privacy while maintaining face structure intelligibility. The different block sizes allow better adaptation to different facial features and regions.
Data Source
AI summary
A method and system for image processing of an image captured by an image-capturing device, can involve dividing a region of interest in an image into blocks, wherein the image is captured by an image-capturing device. The region of interest can be associated with one or more faces in the image, and a resulting division of the dividing of the region of interest may be non-uniform. A minimum block size can be defined and the coordinates of each block after dividing of the region of interest in the image into the blocks. The one or more faces in the image can be redacted and reconstructed using the coordinates. In addition, redacting and reconstructing of the image is reversible for use with detecting the one or more faces in the image. The coordinates can specify top left coordinates and bottom right coordinates of each block.


