Image analysis – Applications
Reexamination Certificate
2000-08-30
2002-08-06
Johns, Andrew W. (Department: 2621)
Image analysis
Applications
C713S176000
Reexamination Certificate
active
06430301
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates to a system for embedding and detecting digital watermarks.
Digital watermarks have been proposed and used for copyright protection of signals such as audio, video, and images. The primary objective is to hide an auxiliary signal within a host signal in such a way that it is substantially imperceptible and difficult to remove without damaging the host signal. The auxiliary signal may carry some information that is helpful, e.g., in a copyright protection mechanism. For example, it can communicate “no copy allowed” to a compliant copying device, and/or it can carry a signature code that identifies the rightful owner, author, and/or content. The signature code can be used to monitor usage of copyrighted material, resolve ownership disputes, keep track of royalties, etc.
Further, digital watermarks can be used to distinguish different copies of the same host signal distributed to different users in legitimate transactions. The primary objective is, if a copy is pirated and redistributed illegally, to trace down the user who obtained it in the legitimate transaction and, presumably, to prosecute him for breach of copyright laws. The secondary objective is to deter casual copying, e.g., among small groups of people.
For example, a pirate can order music over the Internet from a legitimate distributor, directly or using a proxy. Then, the pirate can resell it or redistribute it for free using the Internet or other means. A similar scenario can occur in the distribution of video or other images (e.g., still photos, computer graphics and games, etc.) over the Internet, or in the distribution of video or music over “pay-per-view” channels in a cable or satellite TV network.
Moreover, in the Internet distribution business, the host signal (music, image or video) is typically stored and delivered in a compressed form (e.g., MP3 format in music). This means that a typical watermark embedding process requires decompression, embedding, and then recompression before transmission. Obviously, this imposes additional processing requirements, and adds more noise to the host signal in the process.
Furthermore, many watermark embedding processes are subject to collusion attacks. A major distinction between Transaction Code Embedding (TCE), also sometimes referred to as “fingerprinting”, and embedding of other messages, such as content ID, owner ID, copy control codes, etc., is that, with TCE, a pirate can design special kind of attacks based on the fact that TCE embeds different auxiliary signals into the same content.
For example, by simply subtracting two copies with different watermarks, the pirate obtains the difference of the pure watermarks, which can help him analyze the hiding technique and devise a sophisticated attack. Secondly, the pirate can average a number of copies to weaken individual watermarks, make them interfere, and eventually make them undetectable. Similarly, the pirate can cut portions of different copies and splice them together. The resulting signal has segments of different watermarks spliced together, which is hard to use to identify the pirate.
One existing technique to fight collusion attacks, described in International Publication no. WO 99/39344, published Aug. 5, 1999 to J. M. Winograd et al., is based on a random phase-modulation technique that precedes the watermark embedding stage. The random phase modulation, although imperceptible to a user, makes copies of the same original sufficiently different so that the collusion attack cannot work. For example, averaging multiple copies of a piece of music produces phase cancellations that make the resulting signal annoying to the listener.
However, this technique does not address the issue of the processing (computational) load of TCE. To the contrary, it proposes an additional processing stage that can only increase the processing load.
Another technique, described in International Publication no. WO 99/62022, published Dec. 2, 1999, to D. Wong and C. Lee, greatly reduces the real-time processing required for TCE by preprocessing a host signal to provide two uncompressed copies, one containing segments with an embedded binary “0”, while the other contains corresponding segments with embedded binary “1”. Successive segments are selected from one of the two copies to provide a time-multiplexed composite host signal with an embedded binary data that corresponds to the transaction code.
However, this technique does not address the security issue and collusion attacks. Moreover, splicing of the segments may result in perceptible artifacts. Additionally, this technique does not address the issue of combining two copies that are saved in a compressed form.
Accordingly, it would be desirable to provide a watermark embedding and detection system that addresses the above and other concerns.
The system should not require decompression and recompression. It is also desirable that the same technique can be applied to different compression/decompression algorithms (such as MPEG, AAC, AC3, ATRAC, etc. in music).
Furthermore, the system should not be overly computationally intensive and costly since embedding is performed frequently (into every copy, not into every original). Although some complex algorithms can make sense for embedding a high quality, high security watermark in a production studio, it may well be too costly to run it on the fly in the Internet distribution of copyrighted content. The embedding and detection system should not be too costly for such applications.
Moreover, the system should thwart collusion attacks, and should enable identification of an illegitimate distributor of protected content, or, more precisely, the original recipient of the content. Additionally, the system should be easily implementable in Internet distribution and other distribution modes.
Also, the system should avoid perceptible artifacts.
The present invention provides a transaction code embedding and detection system that provides the above and other advantages.
SUMMARY OF THE INVENTION
The present invention relates to a system for embedding and detecting digital watermarks.
Each copy of content to be protected is labeled with a unique code referred to as a “transaction code”. Using the transaction code, it is possible to identify the user that obtained a legitimate copy of the content but distributed it illegally, and prosecute him, or at least blacklist him to prevent his further purchases. Thus, legal action can be taken against the user even when the distributor to the user is immune from legal action, e.g., due to being a foreign-based company.
In one aspect of the invention, a method for embedding watermarks in a host signal, includes the step of forming watermarked copies of the host signal with at least one different transaction watermark and at least one common watermark embedded therein. The host signal may be a music piece (e.g., song) that is to be protected. Thus, a given copy contains transaction watermarks with the same symbol values. Portions of the different watermarked copies are assembled (e.g., using multiplexing) according to a transaction code to form an output signal with transaction watermarks that correspond to the transaction code.
The output signal is then distributed to a user, who can be subsequently identified if the content is re-distributed in an unauthorized manner.
In a further aspect of the invention, a method is presented for analyzing a watermarked signal e.g., which is suspected to have been re-distributed illicitly and modified using a collusion attack. In possible collusion attacks, different copies are cut-and-spliced together or averaged. The method includes the step of recovering a plurality of transaction watermarks of the watermarked signal that define respective symbols thereof. At least one hypothesis transaction code is provided that defines respective symbols thereof. The symbols of the watermarked signal are compared to corresponding symbols of the hypothesis transaction code to determine a correspondence therebetween. Based on t
Johns Andrew W.
Lipsitz Barry R.
Verance Corporation
LandOfFree
Formation and analysis of signals with common and... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Formation and analysis of signals with common and..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Formation and analysis of signals with common and... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2946355