Related Experiment Video
Updated: Apr 7, 2026
12:39
A Methodology for Capturing Joint Visual Attention Using Mobile Eye-Trackers
Published on: January 18, 2020
8.3K
Privacy-Aware Meta-Optics for Person Detection
Zaid Tasneem1, Yongyi Zhao1, Johannes E Fröch2
1Department of Electrical and Computer Engineering, Rice University, Houston, Texas 77005, United States.
ACS Photonics
|April 6, 2026
Summary
This study introduces privacy-aware meta-optics (MOs) that optically filter identity information, enabling person detection while preserving privacy. This computational camera offers a superior privacy-utility balance for computer vision applications.
Area of Science:
- Optics
- Computer Vision
- Computational Imaging
Background:
- Growing privacy concerns due to widespread computer vision use.
- Need for methods that balance data utility with personal privacy.
Purpose of the Study:
- To develop a computational camera that optically filters private attributes like identity.
- To enable downstream vision tasks such as person detection while enhancing privacy.
Main Methods:
- Replacing traditional lenses with broadband meta-optics (MOs).
- End-to-end optimization of MO parameters using differentiable look-up tables and neural networks.
- Implementing a novel private Strehl integral regularization for privacy preservation.
Main Results:
- Experimental validation of privacy-aware meta-optics.
- Demonstrated superior privacy-utility trade-off compared to existing techniques.
- Successful filtering of high-frequency facial identity information while preserving low-frequency details.
Conclusions:
- The first demonstration of privacy-aware broadband meta-optics for person detection.
- A computationally inexpensive and effective method for enhancing privacy in computer vision.
- Significant potential for applications requiring both surveillance and privacy protection.
