Download - De Identification Seminar
![Page 1: De Identification Seminar](https://reader031.vdocument.in/reader031/viewer/2022032015/563db8ac550346aa9a95d9b9/html5/thumbnails/1.jpg)
Person De-Identification
in Videos
![Page 2: De Identification Seminar](https://reader031.vdocument.in/reader031/viewer/2022032015/563db8ac550346aa9a95d9b9/html5/thumbnails/2.jpg)
Guided By :-Ms. Mamata NayakAsst. Professor
Submitted By :-NikitaRegd. No. – 1241017021Csit ‘C’
Department of Computer Science and Information TechnologyInstitute of Technical Education and Research,
BBSR Siksha ’O’ Anusandhan University
![Page 3: De Identification Seminar](https://reader031.vdocument.in/reader031/viewer/2022032015/563db8ac550346aa9a95d9b9/html5/thumbnails/3.jpg)
Proposed Approach
Recognition vs de-identification
De-Identification: General Framework
Conclusions
Outline :
Introduction
![Page 4: De Identification Seminar](https://reader031.vdocument.in/reader031/viewer/2022032015/563db8ac550346aa9a95d9b9/html5/thumbnails/4.jpg)
WHY De-Identification ??
Videos over the internet invades our privacy. HOW?? E.g. Google Street View
As the number of video surveillance systems increases, ensuring privacy gets important.
![Page 5: De Identification Seminar](https://reader031.vdocument.in/reader031/viewer/2022032015/563db8ac550346aa9a95d9b9/html5/thumbnails/5.jpg)
WHY De-Identification ??
May exist the need to see the individuals in them, identifying the action suffices in most cases.
The actor needs to be identified only rarely and only to authorized personnel.
Hence, De-Identification !!
![Page 6: De Identification Seminar](https://reader031.vdocument.in/reader031/viewer/2022032015/563db8ac550346aa9a95d9b9/html5/thumbnails/6.jpg)
Recognition vs. De-Identification: Recognition & de-identification are opposite to each other.
Recognition De-identification
Reco
gn
itio
n
De-Id
entifi
catio
n
makes use of all possible features to identify a person.
prevents a person’s identity from being connected with information to thwart recognition
![Page 7: De Identification Seminar](https://reader031.vdocument.in/reader031/viewer/2022032015/563db8ac550346aa9a95d9b9/html5/thumbnails/7.jpg)
De-Identification: General Framework Easy to hide the identity of individuals by replacing
a conservative area around them by, say, black pixels.
Goal is to protect the privacy of the individuals while
providing sufficient feel for the human activities in the
space being imaged.
Privacy protection provided should be immune to recognition
using computer vision as well as using human.
![Page 8: De Identification Seminar](https://reader031.vdocument.in/reader031/viewer/2022032015/563db8ac550346aa9a95d9b9/html5/thumbnails/8.jpg)
A. Different Scenarios & De-identificationThree types of videos:
1. Casual videos: captured for other purposes and gets shared.
2. Public Surveillance videos: come from cameras watching spaces such as airports, streets and so on.
3. Private surveillance videos: cameras placed at the entrances
of semi-private spaces like offices.
![Page 9: De Identification Seminar](https://reader031.vdocument.in/reader031/viewer/2022032015/563db8ac550346aa9a95d9b9/html5/thumbnails/9.jpg)
B. Criteria for De-identificationFeatures to recognize humans:
1. Face plays a dominant role in automatic and manual
identification.
2. The body silhouette and the gait are important clues
available in videos.
3. Race and gender, hard to mask completely.
![Page 10: De Identification Seminar](https://reader031.vdocument.in/reader031/viewer/2022032015/563db8ac550346aa9a95d9b9/html5/thumbnails/10.jpg)
C. Subverting De-identification
de-identification can be “attacked” to reveal the identity of individuals involved
1. Reversing the de-identification transformation.
2. Recognizing persons from face, body outline, gait etc.
![Page 11: De Identification Seminar](https://reader031.vdocument.in/reader031/viewer/2022032015/563db8ac550346aa9a95d9b9/html5/thumbnails/11.jpg)
C. Subverting De-identification (contd.)de-identification can be “attacked” to reveal the identity of individuals involved
3. Manual identification is another way to subvert
de-identification, though it is considerably more
expensive.
4. Brute-Force verification is another way to attack.
![Page 12: De Identification Seminar](https://reader031.vdocument.in/reader031/viewer/2022032015/563db8ac550346aa9a95d9b9/html5/thumbnails/12.jpg)
D. Storage of Videos1. The de-identification should be selectively reversed
when needed. 2. Safest approach is to de-identify the video at the capture-camera. Only the transformed video is recorded.
3. Another approach is to store the original video, with
sufficiently hard encryption, along with the de-identified
video.
![Page 13: De Identification Seminar](https://reader031.vdocument.in/reader031/viewer/2022032015/563db8ac550346aa9a95d9b9/html5/thumbnails/13.jpg)
De-identification Model
The system is comprised of three modules:
1. Detect & track2. Segmentation
3. De-Identification
![Page 14: De Identification Seminar](https://reader031.vdocument.in/reader031/viewer/2022032015/563db8ac550346aa9a95d9b9/html5/thumbnails/14.jpg)
A. Detect & Track
1. The first step is to detect the presence of a person in the scene.
2. Patch-based recognition approach for object tracking.
3. To avoid errors resulting from fast changing scale, we apply the human detector every ‘F’ frames
4. The output of the human detector becomes the input to the
tracking module. The value of F depends on the amount
of movement in the video.
![Page 15: De Identification Seminar](https://reader031.vdocument.in/reader031/viewer/2022032015/563db8ac550346aa9a95d9b9/html5/thumbnails/15.jpg)
B. Segmentation
1. The faces of the human in every frame, are stacked across time to generate a video tube of the person.
2. Multiple video tubes are formed if there are multiple people in the video.
3. Rigid but blocky outline of the human is obtained after segmentation.
![Page 16: De Identification Seminar](https://reader031.vdocument.in/reader031/viewer/2022032015/563db8ac550346aa9a95d9b9/html5/thumbnails/16.jpg)
C. De-identification
Two de-identification transformations:
1. Exponential blur of pixels.
2. Line Integral Convolution (LIC).
![Page 17: De Identification Seminar](https://reader031.vdocument.in/reader031/viewer/2022032015/563db8ac550346aa9a95d9b9/html5/thumbnails/17.jpg)
Exponential blur of pixels
Blurring of the segmented area by continuous overlapping till person’s identity gets hidden.
Suitable only where gait is not involved.
![Page 18: De Identification Seminar](https://reader031.vdocument.in/reader031/viewer/2022032015/563db8ac550346aa9a95d9b9/html5/thumbnails/18.jpg)
Conclusion Presented a basic system to protect privacy against human recognition
Issues relating to de-identification of individuals in videos to protect their privacy were analyzed
Characteristics are difficult to hide if familiarity is high to the user
Blurring is a good way to hide the identity if gait is not involved.
![Page 19: De Identification Seminar](https://reader031.vdocument.in/reader031/viewer/2022032015/563db8ac550346aa9a95d9b9/html5/thumbnails/19.jpg)
Bibliography
[1] “De-Identification.” Wikipedia: The free Encyclopedia. Wikipedia Foundation, Inc. 12th June 2015. Web. 23 Jul 15. < https://en.wikipedia.org/wiki/De-identification>
[2] P. Agrawal and P. J. Narayanan, “Person De-Identification in Videos”, IEEE Transactions on circuits and systems for video technology, Vol. 21, No. 3, pp. 299-310, Mar. 2011
![Page 20: De Identification Seminar](https://reader031.vdocument.in/reader031/viewer/2022032015/563db8ac550346aa9a95d9b9/html5/thumbnails/20.jpg)
![Page 21: De Identification Seminar](https://reader031.vdocument.in/reader031/viewer/2022032015/563db8ac550346aa9a95d9b9/html5/thumbnails/21.jpg)