This paper describes our participation in the Visual Privacy Task of MediaEval 2014, which aims to obscure human occurrence in image sequences. As a result the recorded person should be unrecognisable, but if needed the obscured areas can be recovered. We use an approach which models the background and pseudo-randomly scrambles pixels within disjunct foreground areas. This technique is reversible and preserves the colour characteristic of each area. So, colourbased approaches will still be able to automatically distinguish between di erently dressed individuals. The evaluations of our results show that the privacy aspect got a high score in all three evaluation streams. The level of intelligibility and the pleasantness of our approach is below the average, since scrambling results in lower `aesthetic' images.
1 Communication Systems Group Technische Universität Berlin, Germany
Video surveillance of public spaces is expanding.
Consequently, individuals are increasingly concerned about the
`invasiveness' of such ubiquitous surveillance and fear that
their privacy is at risk. The demands of stakeholders to
prevent criminal activities are often seen to be in con ict with
the privacy requirements of individuals. The main challenge
is to preserve the anonymity of the surveyed individuals and
also to ful l the stakeholders needs. The problem of privacy
protection in video surveillance is concerned in this year's
MediaEval Visual Privacy Task [
Our proposed privacy-protection approach consists of a
background modelling module and a scrambling module that
obfuscates foreground masks. Since the PEViD videos [
We use background subtraction for generating a foreground
mask for each frame. In order to compensate slight camera
movements, each frame is subsampled by two and the
resulting masks are interpolated properly. Our background
modelling module relies on a improved background
subtraction scheme [
These foreground areas are then obfuscated by shu ing their pixels. So, an obfuscated area di ers from its original version in a changed sequence of their pixels.
The shu e algorithm is based on a modi ed variant of the
Fisher-Yates method [
So, the permutation of the pixels of each foreground areas is determined by the order generated by a pseudo-random sequence. The pseudo-random sequence is repeatable due to the characteristics of the pseudo-random number generator (PRNG). The PRNG produces a random, but repeatable sequence of integer numbers by specifying a certain, but xed seed. This seed is generated from the hash value of a chosen password. This value is xed for all regions in each frame and video sequence. Since the pseudo-random sequence is repeatable through the given seed, the permutation of pixels is reversible. So, the scrambled image regions can be recovered by knowing the password and the shape of each disjunct scrambled area. We choose for scrambling instead of cryptography to be robust against image compression artefacts and transmission errors. Those errors will also a ect the recovered frame in terms of distorted pixels, but these errors will not break the de-scrambling scheme.
The video sequences of the VPT dataset [
The evaluation of the obscured videos took place using subjective procedures. Three di erent groups are asked to survey the videos and respond to question concerning the content (number of persons, actions, etc. ). Three metrics are generated from these surveys: pleasantness, intelligibility, and privacy. These groups contains of crowdsourced workers and two focus groups, the scores based on their opinions is shown in Table 1.
Pleasantness stands for the in uence of the obscuring lter on the human perception of the image distortion. The subjective score is based on the level of user acceptance. Here the score is below the median value resulting from distraction of the users.
Intelligibility stands for the ability of identifying actions and objects within video frames. All three groups evaluate our lter with high scores that are close to the median. Since full masks of person are retained, their action should be recognizable.
The privacy metric concerns about the identi cation of individuals through their faces, ethics or personal accessories. This score is much higher the average. A high subjective score was excepted, since it is very hard for the human eye to recognise structures within scrambled areas.
We expect higher score in all three categories when applying a more accurate background subtraction algorithm. 4.
We propose a reversible approach for scrambling foreground masks within images or videos to obscure its content. This approach ensures a high level of privacy, and achieves a standard level in the other aspects, like pleasantness and intelligibility. In future we will investigate the e ect of more accurate foreground masks on these privacies scores. The clue is that these areas can be recovered for further analysis, if the foreground mask and the password which generated the seed for the pseudo-random number generation are known. 5.
The research leading to these results has received funding from the European Community's FP7 under grant agreement number FP7-261743 (VideoSense). 6.