Intellectual property rights protection and management of multimedia data is essential for the deployment of e-commerce systems involving transactions on such data. Lately, replica detection or fingerprinting has emerged as a promising approach for the rights management of multimedia data. In this chapter, a review of 2 replica detection techniques is presented. The first technique utilizes color-based descriptors, an R-tree indexing structure, and Linear Discriminant Analysis (LDA) to achieve image replica detection. The second technique is a video fingerprinting method that utilizes information about the appearances of individuals in videos along with an efficient search and matching strategy.
Numerous systems addressing the issue of copyright protection and DRM in general can be found in the literature, the vast majority of them being based on watermarking. Watermarking is the technique of imperceptibly embedding information within a medium (Tefas et al. 2005). Although watermarking has been the subject of intensive research in both the industry and the academia, it has certain disadvantages. Embedding information inside a multimedia item before it becomes available to the public, implies that the data will be distorted up to a certain extent and that watermarking methods are not applicable to data that are already in the public domain and need to be protected. Moreover, watermarking is unable to cope with leakage of unprotected content, i.e., cases where a copy of the original item that bears no watermark is stolen and distributed.
Recently, the scientific community started to investigate digital rights management in multimedia data from an alternative point of view i.e. as a problem of similarity of such data, the similarity being defined in a robust way. These approaches, which come under different names, such as multimedia fingerprinting (Oostveen et al. 2002), robust or perceptual hashing (Michak & Venkatesan 2001), replica or near-replica recognition/detection (Ke et al. 2004) and copy detection (Kim & Vasudev 2005) aim at extracting from the data a feature vector, called perceptual hash, fingerprint or signature, that characterizes them in a unique, robust and discriminative way. This feature vector can be combined with a database of multimedia documents that need to be managed with respect to their digital rights, an appropriate similarity metric and an efficient database search strategy in order to devise a DRM system. More specifically, such a system can decide if a query digital item resembles a reference item in the database. If this is indeed the case, the query item is identified as being a copy (replica) of the corresponding item in the database and legal action can be pursued against its owner/distributor if he is not legally possessing/distributing it. In order to be of practical use, the feature vectors and the matching procedure involved in a fingerprinting system should be robust to manipulations that multimedia data might undergo, either due to their distribution and use or due to an intentional attempt to make them unrecognizable by the fingerprinting system. Unlike watermarking, no information needs to be embedded within the multimedia content in a fingerprinting system, thus ensuring perfect quality for the data to be protected and furthermore making the system applicable to data that are already in the public domain. It should be mentioned here that the term fingerprinting as used in this chapter and in other papers, should not be confused with the fingerprinting watermarking which is essentially a variant of watermarking.