Compact Video Fingerprinting via Structural Graphical Models

Abstract

Much previous work in video fingerprinting has focused on robustness and security issues, but the compactness requirement, i.e., the hash should be of a short length with acceptable robustness and discriminability, continues to be a significant practical challenge. In this paper, we propose a video fingerprinting method with explicit attention on compactness. First, we develop a new graphical representation of the video which reduces temporal redundancies and makes robust feature extraction much more economical. Second, a randomized adaptive quantizer is proposed to further decrease the final hash length while maintaining acceptable detection performance in terms of receiver operating characteristics (ROCs). Experimental results reveal that the proposed method offers a more favorable robustness versus discriminability tradeoff over the state of the art particularly when the bit budget of the video fingerprint is low.Index Terms-Compact video fingerprint, randomized adaptive quantizer, structural graphical models.

Perceptual Video Hashing for Content Identification and Authentication

“…12-14. The results are in perfect agreement with those reported in [23] in the sense that CGO is significantly outperformed by the TIRI and SGM hashing techniques. It can be seen that CGO cannot withstand the applied attacks.…”

“…12). It has been reported that the strength of SGM lies in its ability to identify the video content when the bit budget of the hash is low [23]. However, with sufficiently large fingerprints, TIRI seems to perform slightly better than SGM according to our experimental results.…”

“…Four well known techniques have been adopted in our comparative study: Centroid of Gradient Orientations (CGO) hashing [11], Temporally Informative Representative Images (TIRI) hashing [20], Weber Binarized Statistical Image Features (WBSIF) hashing [28], and Structural Graphical Models (SGM) based hashing [23]. We have used our own implementation of [11], [28], and [20] and the authors' implementation of [23] which has been available in [45]. The same parameters setting of the aforementioned systems has been adopted here as suggested by their authors.…”

“…Li and Monga have used multi-linear subspace projections with a reduced rank factorization to extract the fingerprint as a summarized version of the video [22]. In a more recent paper, the authors have proposed to use structural graphical models to encode the temporal evolution of the video content [23]. The real-valued hash is then projected onto a randomly generated space whose components are drawn i.i.d.…”

A robust audio fingerprinting method using spectrograms saliency maps

“…In this way, feature extraction methods used over images can be applied in this signal representation. A Saliency Map (SM) is a kind of global feature widely used [12]- [14] that represents the most visual prominent regions of an image [15].…”