Digital Video Watermarking

Project

Short Introduction to Watermarking

History

The roots of watermarking are considered to be in the study of “Steganography”. The word comes from the old Greek language and can be translated as “cover writing”. Steganography was basically a way of transmitting hidden (secret) messages between allies, being used as early as 1000 B.C. First references to steganography appear in Homer’s “Iliad” and “Histories of Herodotus”(440 B.C.)

Definition

Generally by watermarking one is hiding a message signal into a host signal, without any perceptual distortion of the host signal. As the word “watermarking” suggests, the mark itself is “transparent” or unnoticeable for the human perception system. Usually, the host signal is a digital media, like audio, video or images. As we all know, the Human Visual System (HVS), is far from being perfect and for images/video it is possible to modify the pixel values without the watermark being visible. Providing that a certain HVS threshold is not exceeded, the modified (watermarked) image/video will be undistinguishable to the human eye compared with the original.

Applications of digital watermarking

The main application of digital watermarking is in copyright protection. The owner of the image/video adds a watermark to his material before it is distributed. In this way is possible to track illegal copies of the copyrighted material.

Possible applications are:

- Broadcast monitoring of video sequences (digital TV)

- DVD protection and access control

- Database retrieval

- Robust identification of digital content

Classification of watermarks

Excluding the obvious case of visible watermarks, we can classify the watermarks as fragile or robust. The fragile watermark is used for detecting even the smallest alteration of an image, while the robust one is specially designed to withstand a wide range of “attacks”, which basically are trying to remove the watermark, but without destroying the image/video.

A watermark can be added to the uncompressed data (raw data), such as a standard uncompressed video sequence as described by ITU-R 601, or can be added to the compressed bit-stream (MPEG2).

The easiest (simplest) way to watermark an image/video, is to change directly the values of the pixels, in the spatial domain. A more advanced way to do it, is to insert the watermark in the frequency domain, using one of the well known transforms: FFT, DCT or DWT. Other techniques are possible as well, like using fractals for example.

The watermark embedding can be done uniformly (or in some other empirical manner), which doesn’t account for the HVS properties (this is called non-perceptual marking). Or, the watermarking embedding can use some HVS models in order to optimise the embedding. Depending on the HVS model used, the perceptual marking can be video independent (basic HVS model) or preferably video dependent (advanced HVS model).

Click me for more … Click me for more …

Last updated: 22 November 2001

Page designed and maintained by Cristian V. Serdean