Video and Photo Anonymization for Media and Journalists – Definition
Video and photo anonymization for media and journalists is a structured process of removing or limiting the ability to identify natural persons in video footage and images intended for publication, archiving, or sharing with third parties. In practice, this mainly concerns faces and license plates visible in the frame. In an editorial context, the goal is not to process documents or text-based data, but to modify visual material so it can be used lawfully and in line with the principle of data minimization.
From the perspective of data protection law and image rights, recording anonymization is an organizational and technical measure that reduces the risk of infringing the rights of people appearing in the material. In the case of photos and videos, identification may result not only from the caption, but from the image itself. As a rule, a face is personal data if it allows a person to be identified directly or indirectly. This approach follows from Article 4(1) of the GDPR, while the obligation to apply appropriate data protection measures and privacy by design arises, respectively, from Article 5(1)(c), Article 25, and Article 32 of Regulation (EU) 2016/679.
In the work of newsrooms and photo agencies, anonymization most often takes the form of blurring or masking faces and license plates before publication. If identification is still possible despite the use of a filter, this cannot be considered effective anonymization. In that case, it is more accurate to speak of an insufficient visual mask rather than anonymization.
Responsibilities of Newsrooms and Photo Agencies When Publishing a Person’s Image
Publishing photos and recordings featuring individuals requires a parallel assessment of personal data protection rules and national laws governing image rights. For editorial teams, an important point is that lawful publication does not always mean anonymization is unnecessary. If a piece of content has high news value but does not require showing a recognizable face, face blurring may be a proportionate solution.
In Poland, the key legal framework includes the GDPR, Articles 23 and 24 of the Civil Code concerning personal rights, and Article 81 of the 1994 Copyright and Related Rights Act. The latter establishes the general rule of consent for disseminating a person’s likeness and sets out the main exceptions.
- a well-known person, where the image was captured in connection with the performance of public functions, in particular political, social, or professional functions,
- a person who constitutes only a detail of a larger whole such as an assembly, landscape, or public event,
- a person who received agreed remuneration for posing, unless they reserved otherwise.
Outside these situations, the editorial team should assess whether publishing a recognizable image is necessary. If it is not necessary, anonymizing the face is a practice consistent with the principle of data minimization. This applies in particular to materials involving crime victims, minors, witnesses, patients, participants in police or emergency interventions, and bystanders.
Photo and Video Anonymization Technologies
Editorial content typically involves two technical stages: object detection and mask application. Automatic face detection and license plate detection are now usually based on machine learning models, including deep learning. The AI model is used to detect the visual features of faces and plates under varying lighting conditions, scales, angles, and image quality. The model itself does not yet anonymize the material. Its role is to identify the area that should be blurred or masked.
In production workflows, object tracking between frames is important so that the mask does not “lose” the face during camera movement or partial occlusion. For still images, localization accuracy is crucial, while for video, mask stability over time is equally important.
Gallio PRO automatically masks only faces and license plates. It does not automatically detect company logos, tattoos, name badges, documents, or content displayed on monitor screens. Such elements can be masked manually using the built-in editor. The software does not perform real-time anonymization or video stream anonymization. It is designed for file-based processing.
Key Parameters and Metrics for Video Anonymization
The effectiveness of video anonymization should not be assessed solely on the basis of an operator’s subjective judgment. For compliance teams and Data Protection Officers, measurable detection quality parameters and the risk of re-identification after publication are what matter most.
Parameter
Practical significance
Typical interpretation
Precision
The proportion of correct detections among all detections
Low precision means too many false masks
Recall
The proportion of detected faces or license plates out of all those present in the material
Low recall means a risk of leaving data unmasked
IoU – Intersection over Union
The degree of overlap between the detection box and the object
It affects whether the mask actually covers the face or plate
Tracking stability
Keeping the mask on the object between frames
Critical for dynamic video footage
File processing time
The time required for analysis and final rendering
It affects newsroom workflow efficiency
When assessing risk, a simple operational rule can be adopted: the higher the recall and the fewer the cases in which a face remains visible even for a single frame, the lower the risk of a breach. In an editorial environment, a false negative is usually a bigger problem than a false positive, because an undetected face creates a possibility of identifying the person.
License Plates and Faces – Scope of Anonymization
In press materials, vehicle registration numbers often appear alongside faces. In some European countries, blurring license plates is treated as a precautionary compliance standard under national law, supervisory authority practice, or editorial policy. In Poland, however, the situation is not entirely uniform.
On the one hand, the Polish Data Protection Authority (UODO), the EDPB, and the case law of the CJEU support a broad interpretation of personal data where identification is possible indirectly. On the other hand, some administrative court rulings suggest that a license plate number alone does not always constitute personal data. In editorial practice, this means the context must be assessed. If a number may lead to identification of the vehicle owner or user, blurring it is a cautious solution consistent with the principle of risk reduction.
Data Security and the On-Premises Model
For newsrooms processing sensitive material, not only the visual outcome matters, but also the way files are processed. On-premises software limits the transfer of recordings outside the organization and makes it easier to meet the security requirements of Article 32 GDPR. This is especially important for investigative materials, footage from crime scenes, and content involving minors or confidential sources.
Gallio PRO operates as on-premises software. In addition, it does not collect logs containing face detection or license plate detection data, and it does not store logs containing personal data or special category data. From the perspective of a Data Protection Officer, this reduces the risk surface and limits the scope of operational data that would otherwise require additional safeguards.
Legal References and Sources
The definition and practice of anonymizing recordings for media should be based on legal acts and guidance, not on editorial custom alone. The following sources are directly relevant or provide interpretative guidance for the publication of photos and videos.
- Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 – GDPR, in particular Articles 4, 5, 25, and 32
- Act of 23 April 1964 – Civil Code, Articles 23 and 24
- Act of 4 February 1994 on Copyright and Related Rights, Article 81
- EDPB Guidelines 4/2019 on Article 25 GDPR – Data Protection by Design and by Default, finally adopted in 2020
- CJEU case law concerning the broad interpretation of identifiability and the practice of supervisory authorities regarding visual data
See Also
- Face blurring
- License plate blurring
- GDPR compliance in video and photo anonymization
- On-premises software for visual data anonymization