Abstract
Digital rights management (DRM) of depth-image-based rendering (DIBR) 3D video is an emerging area of research. Existing schemes for DIBR 3D video cause video distortions, are vulnerable to severe signal and geometric attacks, cannot protect 2D frame and depth map independently or can hardly deal with large-scale videos. To address these issues, a novel zero-watermark scheme based on invariant feature and similarity-based retrieval for protecting DIBR 3D video (RZW-SR3D) is proposed in this study. In RZW-SR3D, invariant features are extracted to generate master and ownership shares for providing distortion-free, robust and discriminative copyright identification under various attacks. Different from traditional zero-watermark schemes, features and ownership shares are stored correlatively, and a similarity-based retrieval phase is designed to provide effective solutions for large-scale videos. In addition, flexible mechanisms based on attention-based fusion are designed to protect 2D frame and depth map independently and simultaneously. Experimental results demonstrate that RZW-SR3D have superior DRM performances than existing schemes. First, RZW-SR3D can extracted the ownership shares relevant to a particular 3D video precisely and reliably for effective copyright identification of large-scale videos. Second, RZW-SR3D ensures lossless, precise, reliable and flexible copyright identification for 2D frame and depth map of 3D videos.
Introduction
As 3D video technology proliferates, protecting intellectual property in Depth-Image-Based Rendering (DIBR) formats has become a critical challenge. DIBR pairs a standard 2D video frame with a grayscale depth map to synthesize 3D views. Traditional watermarking methods often fall short: embedding data directly into 2D frames causes visible distortion, while watermarks in depth maps are frequently destroyed by signal processing or geometric attacks like rotation and cropping.
Zero-watermarking offers a solution by generating a cryptographic "share" based on the video's inherent features rather than altering the pixels. However, traditional zero-watermarking lacks a search mechanism; identifying a pirated copy in a large database requires knowing exactly which original video to compare it against. The proposed RZW-SR3D (Robust Zero-Watermark with Similarity-Based Retrieval) addresses this by integrating a retrieval system that can scan massive databases to identify pirated content without ever modifying the original video file.
Proposed Methodology
The RZW-SR3D framework operates through three distinct phases designed to ensure robustness and scalability.
1. Feature Extraction and Registration
The system extracts invariant features from both the 2D frame and the depth map. It utilizes Temporally Informative Representative Images (TIRI) to capture temporal dynamics and partitions frames into concentric rings. This geometric approach ensures the features remain stable even if the video is rotated or cropped. Using (2,2) Visual Secret Sharing (VSS), these features are converted into a "master share," which is combined with a watermark to create an "ownership share" stored in a database.
2. Similarity-Based Retrieval
When a suspicious video is queried, the system extracts its features and performs a search across the database using normalized distances. A key innovation is the attention-based fusion mechanism, which intelligently weights the search results from the texture-rich 2D frame and the structurally stable depth map to improve accuracy.
3. Copyright Identification
If a match is found during retrieval, the system retrieves the corresponding ownership share. By stacking the queried video's master share with the database's ownership share, the original watermark is visually revealed, proving ownership without having distorted the original media.
Experimental Evaluation
The researchers tested the scheme on a dataset of 200 DIBR 3D video clips, generating a testing pool of 32,000 sequences subjected to 14 types of attacks (e.g., Gaussian blurring, frame dropping, and rotation).
- Retrieval Accuracy: The attention-based fusion of 2D and depth features achieved an exceptionally low average false-negative rate ($P_{fn}$) of 0.0029, significantly outperforming single-modality searches.
- Robustness: The system maintained a low Bit Error Rate (BER) of 0.0287 for recovered watermarks under attack, far superior to existing state-of-the-art methods like AUEW (0.1759) and RZW3D (0.1185).
- DIBR Synthesis: The scheme remained effective even on newly synthesized 3D views, which involve horizontal pixel shifts, maintaining a BER of approximately 0.04.
Conclusion
RZW-SR3D provides a distortion-free, robust, and scalable solution for 3D video copyright protection. By bridging the gap between zero-watermarking and large-scale retrieval, it enables practical piracy detection in massive databases. Future research will explore applying this lossless technique to medical imaging (e.g., 3D CT or MRI scans), where maintaining data integrity is a clinical necessity.
Summary updated