Digital Fingerprints for Videos
IEEE Access · 2026 · 14 · 100810-100825
DOI: 10.1109/access.2026.3707469 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Zifu Qin und 2 weiteren Autor:innen. Die Arbeit nennt 43 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Cornell University.
Abstract (Originalsprache)
Millions of videos are uploaded daily across platforms such as TikTok, YouTube, LinkedIn, Netflix, and Facebook. Yet many systems still rely on externally assigned identifiers, which make content-based search, duplicate detection, and copyright enforcement difficult when videos undergo content-preserving modifications. This paper introduces an objective method that creates a unique digital fingerprint for each distinct video. It presents the Structural Similarity Fingerprinting (SSF) framework, a computationally lightweight and theoretically grounded video fingerprinting framework built on the Structural Similarity Index (SSIM). The method quantizes temporal SSIM observations into compact byte-level fingerprints that capture intrinsic structural content, enabling fast matching and scalable indexing. We evaluated the framework under geometric flipping, a common and challenging content-preserving transformation, using 750 source videos and 2,250 flipped variants across three orientations (horizontal, vertical, and combined axes). In these experiments, under controlled conditions, the proposed fingerprints matched exactly and achieved 100% continuous matching, substantially outperforming conventional perceptual hashing baselines, which reached at most 68.3%, and exceeding the flip-robust RFR-35 method under stricter Hamming distance thresholds. The framework ran up to $4.33\times $ faster than standard perceptual hashing methods for extraction and retrieval, while requiring approximately $5.9\times $ less storage than 64-bit-per-frame representations. These results demonstrate that SSIM-based structural fingerprints provide an efficient foundation for robust video identification in large-scale retrieval, registration, and copyright workflows.
Förderung
- Cornell University
Zitieren
APA
Qin, Z., Misra, U., & Srivastava, M. (2026). Digital Fingerprints for Videos. IEEE Access, 14, 100810-100825. https://doi.org/10.1109/access.2026.3707469
Vancouver
Qin Z, Misra U, Srivastava M. Digital Fingerprints for Videos. IEEE Access. 2026;14:100810-100825. doi:10.1109/access.2026.3707469
BibTeX
@article{Qin2026,
author = {Qin, Zifu and Misra, Utkarsh and Srivastava, Madhur},
title = {Digital Fingerprints for Videos},
journal = {IEEE Access},
year = {2026},
volume = {14},
pages = {100810--100825},
doi = {10.1109/access.2026.3707469}
}
EUVIMED