Capacity, robustness, and fidelity are the most crucial parameters that govern an efficient video-watermarking solution for the OTT space. Even if the OTT content is protected and managed through digital rights management (DRM) technology, there is always a trade-off among these factors that determine how effectively the content producer can identify the rogue element that leaks their content.
Capacity or payload is essentially the amount of information (given in number of bits/sample) that can be stored in a forensic watermark, which is crucial to identify the source of piracy of premium content. Often, the digital payload of a watermark depends on the size of host data, which is the video asset. Robustness, on the other hand, is the ability of a watermark-detecting algorithm to extract the watermarking data after it has been altered or tampered with by pirates. Therefore, robustness can be evaluated by the survival of the watermark after attacks. Fidelity, also known as imperceptibility, is the ability of a watermark to not be detected by a human observer. The introduction of suspicious perceptible artifacts into the DRM protected content can help attackers in detecting the precise location of the watermark, thereby letting them distort, substitute, or remove the watermark data maliciously. Fidelity can, thus, be evaluated by the similarity between the watermarked and un-watermarked data.
The parameters listed above are all inter-connected and conflicting. For instance, the robustness of a watermark can be increased by making large modifications to the host data for each bit. However, this will be noticeable and would limit the digital payload. Similarly, the data payload can be increased by reducing the number of samples allocated to each hidden bit but this might affect the robustness of the solution. Hence, depending on the requirements of the application, a trade-off has to be found to develop an optimal watermark. For example, most of the time applications prefer a low-energy watermark signal so that distortions remain imperceptible. But this is not true for high-degrading environments which necessitate the use of strong watermarks that can survive the transmission. On the other hand, some applications depend on the weakness of a fragile watermark to ensure the integrity of the digital data.
Hence, each video watermarking approach must be based on a different trade-off, and it must be ensured that the methods under consideration are evaluated under similar conditions. In other words, if watermarking approaches are being compared in terms of robustness, the algorithm should have the same capacity and perceptibility.