Multimedia Analysis for Content Identification

Multimedia content identification methods create a compact bitstream representation of the underlying content that is robust against common signal processing operations while being sensitive to the content. The robustness and sensitivity of the bitstream representation are conflicting requirements. In this chapter, we examine three issues in the context of achieving the tradeoff between robustness and sensitivity. They are (i) the representation domain for content (spatial, time or transform), (ii) local versus global features in the representation domain, (iii) robust hash of features (the first two of these directly relate to multimedia content analysis). We review the algorithms proposed in literature with these three issues in mind. Finally, we present some applications of content identification technology that exist today in the market and discuss the remaining challenges for future applications.