Investigating the Perception of Facial Anonymization Techniques in 360° Videos

TL;DR

This paper examines three facial anonymization techniques—blocking, blurring, and face-swapping—in 360° videos to understand their impact on perceived realism, anonymization effectiveness, and viewer presence. By comparing screen and HMD viewing and using IPQ-based presence measures, the study finds face-swapping to be the least noticeable and most realistic, though it can introduce artifacts and raise privacy concerns; blocking is highly disruptive and reduces presence, while blurring offers a practical trade-off. The results highlight a privacy-realism trade-off in immersive 360° content and emphasize the need for anonymization methods that preserve scene authenticity while protecting bystanders’ privacy. The work also discusses ethical considerations and suggests directions for improving anonymization techniques and for validating actual privacy protection beyond perceptual assessments.

Abstract

In this work, we investigate facial anonymization techniques in 360° videos and assess their influence on the perceived realism, anonymization effect, and presence of participants. In comparison to traditional footage, 360° videos can convey engaging, immersive experiences that accurately represent the atmosphere of real-world locations. As the entire environment is captured simultaneously, it is necessary to anonymize the faces of bystanders in recordings of public spaces. Since this alters the video content, the perceived realism and immersion could be reduced. To understand these effects, we compare non-anonymized and anonymized 360° videos using blurring, black boxes, and face-swapping shown either on a regular screen or in a head-mounted display (HMD). Our results indicate significant differences in the perception of the anonymization techniques. We find that face-swapping is most realistic and least disruptive, however, participants raised concerns regarding the effectiveness of the anonymization. Furthermore, we observe that presence is affected by facial anonymization in HMD condition. Overall, the results underscore the need for facial anonymization techniques that balance both photo-realism and a sense of privacy.

Figures (5)

• Figure 1: In this paper, we investigate three different facial anonymization techniques for 360° videos: a) Blocking of faces with black boxes, b) Blurring the facial area, c) Face-Swapping using a synthetically generated face.
• Figure 2: Exemplar stimuli of different scenes with varying amounts of bystanders (Left: Block, Middle: Blur, Right: Swap).
• Figure 3: A synthetic face (left) is used to swap all faces in the 360° videos. While this produces high-quality results for different people and minor occlusions (middle) artifacts can occur if large areas of the face are occluded and in profile views (right).
• Figure 4: Impression of the anonymization techniques in HMD and Screen condition. Left: Average ratings for trials correctly reported to be anonymized. Right: Answers for the comparison of techniques in the post experiment questionnaire. Error bars indicate the standard error of the mean (SEM).
• Figure 5: Mean ratings for presence (IPQ scales: G - General Presence, INV - Involvement, REAL - Experienced Realism, SP - Spatial Presence) and impressions of the scene in HMD condition. Error bars indicate the SEM.