Privacy Preserving Optics for Miniature Vision Sensors

Similar documents
Sensor-level Privacy for Thermal Cameras

Introduction. Prof. Lina Karam School of Electrical, Computer, & Energy Engineering Arizona State University

Design of Temporally Dithered Codes for Increased Depth of Field in Structured Light Systems

Dappled Photography: Mask Enhanced Cameras for Heterodyned Light Fields and Coded Aperture Refocusing

Computational Cameras. Rahul Raguram COMP

NTU CSIE. Advisor: Wu Ja Ling, Ph.D.

WIDE SPECTRAL RANGE IMAGING INTERFEROMETER

High Resolution Spectral Video Capture & Computational Photography Xun Cao ( 曹汛 )

Midterm Examination CS 534: Computational Photography

Study guide for Graduate Computer Vision

EC-433 Digital Image Processing

Lecture 18: Light field cameras. (plenoptic cameras) Visual Computing Systems CMU , Fall 2013

Demosaicing and Denoising on Simulated Light Field Images

Coded photography , , Computational Photography Fall 2018, Lecture 14

Implementation of Adaptive Coded Aperture Imaging using a Digital Micro-Mirror Device for Defocus Deblurring

A Mathematical model for the determination of distance of an object in a 2D image

Near-Invariant Blur for Depth and 2D Motion via Time-Varying Light Field Analysis

Dynamically Reparameterized Light Fields & Fourier Slice Photography. Oliver Barth, 2009 Max Planck Institute Saarbrücken

Why learn about photography in this course?

Coded photography , , Computational Photography Fall 2017, Lecture 18

lecture 24 image capture - photography: model of image formation - image blur - camera settings (f-number, shutter speed) - exposure - camera response

6.A44 Computational Photography

Module 3: Video Sampling Lecture 18: Filtering operations in Camera and display devices. The Lecture Contains: Effect of Temporal Aperture:

ROBOT VISION. Dr.M.Madhavi, MED, MVSREC

A moment-preserving approach for depth from defocus

ME 6406 MACHINE VISION. Georgia Institute of Technology

Light field sensing. Marc Levoy. Computer Science Department Stanford University

La photographie numérique. Frank NIELSEN Lundi 7 Juin 2010

Computational Approaches to Cameras

Dr F. Cuzzolin 1. September 29, 2015

Overview. Pinhole camera model Projective geometry Vanishing points and lines Projection matrix Cameras with Lenses Color Digital image

Camera Requirements For Precision Agriculture

Unit 1: Image Formation

Ron Liu OPTI521-Introductory Optomechanical Engineering December 7, 2009

Deconvolution , , Computational Photography Fall 2018, Lecture 12

Light-Field Database Creation and Depth Estimation

Privacy Preserving Optics for Miniature Vision Sensors

OFFSET AND NOISE COMPENSATION

LENSLESS IMAGING BY COMPRESSIVE SENSING

Deconvolution , , Computational Photography Fall 2017, Lecture 17

Coding and Modulation in Cameras

VC 16/17 TP2 Image Formation

Spectral and Polarization Configuration Guide for MS Series 3-CCD Cameras

HDR videos acquisition

Exercise questions for Machine vision

IMAGE FORMATION. Light source properties. Sensor characteristics Surface. Surface reflectance properties. Optics

VC 14/15 TP2 Image Formation

Continuous Flash. October 1, Technical Report MSR-TR Microsoft Research Microsoft Corporation One Microsoft Way Redmond, WA 98052

Sensors and Sensing Cameras and Camera Calibration

Computer Vision. The Pinhole Camera Model

Coded Aperture for Projector and Camera for Robust 3D measurement

Camera Requirements For Precision Agriculture

Coded Computational Photography!

PROCEEDINGS OF SPIE. Measurement of low-order aberrations with an autostigmatic microscope

Bias errors in PIV: the pixel locking effect revisited.

Single-Image Shape from Defocus

Digital Image Processing

CHARGE-COUPLED DEVICE (CCD)

Multi-aperture camera module with 720presolution

Acquisition and representation of images

Camera Image Processing Pipeline: Part II


Background. Computer Vision & Digital Image Processing. Improved Bartlane transmitted image. Example Bartlane transmitted image

Depth from Diffusion

Compact Dual Field-of-View Telescope for Small Satellite Payloads

LENSES. INEL 6088 Computer Vision

The ultimate camera. Computational Photography. Creating the ultimate camera. The ultimate camera. What does it do?

Sampling Efficiency in Digital Camera Performance Standards

MINIATURE X-RAY SOURCES AND THE EFFECTS OF SPOT SIZE ON SYSTEM PERFORMANCE

HDR images acquisition

Lenses, exposure, and (de)focus

Introduction. Related Work

New applications of Spectral Edge image fusion

Computer Vision. Howie Choset Introduction to Robotics

CS534 Introduction to Computer Vision. Linear Filters. Ahmed Elgammal Dept. of Computer Science Rutgers University

Image acquisition. In both cases, the digital sensing element is one of the following: Line array Area array. Single sensor

CSE Tue 10/09. Nadir Weibel

VC 11/12 T2 Image Formation

Single Camera Catadioptric Stereo System

Day&Night Box camera with 36x Optical Zoom WV-CZ392/CZ492

Low Cost Earth Sensor based on Oxygen Airglow

Fabrication Methodology of microlenses for stereoscopic imagers using standard CMOS process. R. P. Rocha, J. P. Carmo, and J. H.

Wavefront coding. Refocusing & Light Fields. Wavefront coding. Final projects. Is depth of field a blur? Frédo Durand Bill Freeman MIT - EECS

PLazeR. a planar laser rangefinder. Robert Ying (ry2242) Derek Xingzhou He (xh2187) Peiqian Li (pl2521) Minh Trang Nguyen (mnn2108)

Study of self-interference incoherent digital holography for the application of retinal imaging

TSBB09 Image Sensors 2018-HT2. Image Formation Part 1

Cardinal Points of an Optical System--and Other Basic Facts

Single-shot three-dimensional imaging of dilute atomic clouds

BIG PIXELS VS. SMALL PIXELS THE OPTICAL BOTTLENECK. Gregory Hollows Edmund Optics

High Dynamic Range Imaging

General Imaging System

Sensing Increased Image Resolution Using Aperture Masks

Realistic Image Synthesis

December 28, Dr. Praveen Sankaran (Department of ECE NIT Calicut DIP)

A Review over Different Blur Detection Techniques in Image Processing

SUPER RESOLUTION INTRODUCTION

technology meets pathology Institute of Pathology, Charité Universitätsmedizin Berlin, Berlin, Germany 3 Overview

Image Restoration and Super- Resolution

Optical Engineering 421/521 Sample Questions for Midterm 1

Selection of Temporally Dithered Codes for Increasing Virtual Depth of Field in Structured Light Systems

Transcription:

Privacy Preserving Optics for Miniature Vision Sensors Francesco Pittaluga and Sanjeev J. Koppal University of Florida Electrical and Computer Engineering

Shoham et al. 07, Wood 08, Enikov et al. 09, Agrihouse 15 The next wave of small devices Microrobots Medical devices Remote sensor nodes

In the future, there will be trillions of networked miniature cameras.

Privacy in the Face of Trillions of Eyes Some groups are particularly vulnerable

Our ideas 1) Pre-Capture Privacy Privacy before capture 2) Miniaturizing Algorithms - High performance - Smallest mass and volume We show mobile scale prototypes

Motivating Example We want to: 1) Track/Photograph everyone Group of People 2) Prevent face recognition Quantitative: accurate people tracking and low recognition rate

Block Diagram Light 1. Tracking people with privacy Modulating Optics Light Camera Group of People 2. Photographing people with privacy

Block Diagram Light 1. Tracking people with privacy Modulating Optics Light Camera Group of People 2. Photographing people with privacy

Ray Diagram Display Light Displayed light Camera Group of People Beam splitter

Ray Diagram Display Light Displayed light Group of People Beam splitter Camera Displayed light + Scene radiance

Pre-Capture White-Out Some overexposed pixels Light Group of People Beam splitter Image Captured

Optical K-Anonymity Optically superimpose face Light Group of People Beam splitter Sweeney 2002

Optical K-anonymity Scene Displayed Image Sensor output

Image Formation Model Image pixel Scene point radiance Pixel-radiance map Weight Camera-display transform Optical path split ratio Weight sum of k-1 images

Miniaturization (Orthographic)

Miniaturization (Orthographic) Display M_min (min size for res.) Camera Beam splitter

Miniaturization (Orthographic) Display lmask lbeam M_min (min size for res.) Camera Beam splitter

Volume occupied Display Camera Beam splitter

Miniaturization by translation Display Camera Beam splitter

Miniaturization (Perspective) Please see details in the paper

Block Diagram Light 1. Tracking people with privacy Modulating Optics Light Camera Group of People 2. Photographing people with privacy

Block Diagram Light 1. Tracking people with privacy Modulating Optics Light Camera Group of People 2. Photographing people with privacy

Our key ideas Defocus in time-of-flight (TOF) and thermal domains preserve utility and provide privacy Multiple defocus apertures allow privacy and utility even in visible domains Not effective in RGB for small blur Neustaedter 2006

Privacy Vision Sensor 1 of 3 Light 1. Tracking people with privacy Modulating Optics Light Camera Group of People 2. Photographing people with privacy

Defocused time-of-flight camera Conventional usage With Defocus

Defocused time-of-flight camera Conventional usage With Defocus

Privacy Vision Sensor 2 of 3 Light 1. Tracking people with privacy Modulating Optics Light Camera Group of People 2. Photographing people with privacy

Defocused thermal camera FLIR One with defocus lens

Miniaturization of defocus d u

Miniaturization of defocus d Camera with defocus u

Miniaturization of defocus Camera with defocus

Miniaturization of defocus Angular support Given desired support Given tolerance Camera with defocus Viewing direction Koppal 2013

Miniaturization of defocus Given Our PAMI 2013 u, d Output Defocus parameter and angular res. Biggest feature to anonymize Given

Miniaturization of defocus Given Our PAMI 2013 u, d Output Given Output

Miniaturization of defocus Camera with defocus

Miniaturization of defocus Camera with defocus

Privacy Vision Sensor 3 of 3 Light 1. Tracking people with privacy Modulating Optics Light Camera Group of People 2. Photographing people with privacy

Scale space analysis Gaussian pyramid Lindeberg 1998

Worry about privacy here! Scale space analysis with optical defocus Optical apertures Scene

Scale space detection with optical defocus A classifier is trained on the blobs

Optical Array Miniaturization

Optical Array Miniaturization Optical elements with mass/volume/fov Physical device has size limits

Optical Array Miniaturization Optical elements with mass/volume/fov Physical device has size limits

Optical Array Miniaturization Has two parts Selection Packing We focus on selection Korf et al. 2010

Knapsack Problem These could get added into the design

Optical Knapsack Problem

Optical Knapsack Problem Angular discretization No magic: Pseudo-polynomial approximation

Summary K-anonymity Camera Defocus in thermal and TOF preserves privacy Optical scale space analysis Optical knapsack solution

Future Work: Privacy in Image Formation Materials Image Formation Camera Geometry Acquired Image Lighting

Demo at CVPR 2015 Sensor Setup Pre-Capture White-Out