Use this URL to cite or link to this record in EThOS:
Title: Digital rights management techniques for H.264 video
Author: Ali, Mohammad Athar
ISNI:       0000 0004 2720 6981
Awarding Body: Loughborough University
Current Institution: Loughborough University
Date of Award: 2011
Availability of Full Text:
Access from EThOS:
Access from Institution:
This work aims to present a number of low-complexity digital rights management (DRM) methodologies for the H.264 standard. Initially, requirements to enforce DRM are analyzed and understood. Based on these requirements, a framework is constructed which puts forth different possibilities that can be explored to satisfy the objective. To implement computationally efficient DRM methods, watermarking and content based copy detection are then chosen as the preferred methodologies. The first approach is based on robust watermarking which modifies the DC residuals of 4×4 macroblocks within I-frames. Robust watermarks are appropriate for content protection and proving ownership. Experimental results show that the technique exhibits encouraging rate-distortion (R-D) characteristics while at the same time being computationally efficient. The problem of content authentication is addressed with the help of two methodologies: irreversible and reversible watermarks. The first approach utilizes the highest frequency coefficient within 4×4 blocks of the I-frames after CAVLC en- tropy encoding to embed a watermark. The technique was found to be very effect- ive in detecting tampering. The second approach applies the difference expansion (DE) method on IPCM macroblocks within P-frames to embed a high-capacity reversible watermark. Experiments prove the technique to be not only fragile and reversible but also exhibiting minimal variation in its R-D characteristics. The final methodology adopted to enforce DRM for H.264 video is based on the concept of signature generation and matching. Specific types of macroblocks within each predefined region of an I-, B- and P-frame are counted at regular intervals in a video clip and an ordinal matrix is constructed based on their count. The matrix is considered to be the signature of that video clip and is matched with longer video sequences to detect copies within them. Simulation results show that the matching methodology is capable of not only detecting copies but also its location within a longer video sequence. Performance analysis depict acceptable false positive and false negative rates and encouraging receiver operating charac- teristics. Finally, the time taken to match and locate copies is significantly low which makes it ideal for use in broadcast and streaming applications.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available