Towards a Video Passive Content Fingerprinting Method for Partial-Copy Detection Robust against Non-Simulated Attacks

Area of Science:

• Computer Science
• Digital Signal Processing
• Multimedia Security

Background:

• Passive content fingerprinting is crucial for video identification and monitoring.
• Existing methods struggle with partial-copy detection and robustness against severe transformations due to computational costs and performance limitations.

Purpose of the Study:

• To develop an efficient and robust content fingerprinting method for video identification.
• To address the challenges of partial-copy detection and performance degradation under various video transformations.

Main Methods:

• Extraction of independent binary global and local fingerprints.
• Combination of features for enhanced discrimination against severe transformations.
• Implementation of an efficient multilevel filtering system to accelerate extraction and matching.

Main Results:

• The proposed method demonstrates superior detection scores compared to state-of-the-art techniques on real copied video datasets.
• The fingerprinting approach is effective against signal processing attacks, geometric transformations, and desynchronization.
• The method achieves high granularity, enabling efficient 1-second segment processing for partial-copy detection.

Conclusions:

• The developed content fingerprinting method offers a significant advancement in video identification, particularly for partial copies.
• The approach effectively balances robustness, computational efficiency, and detection accuracy.
• This method is suitable for applications requiring real-time video analysis and large database comparisons.