In this evolving century, all multimedia production and distribution is digital. This surge in digitalization is due to the numerous advantages that the latter provides.
In fact, the most important and commercial advantage of digital media in the domain of creation, processing and distribution is the possibility of easily replicating a digital media (image, video, audio) without any lost of quality.
However, this particular advantage may not be a very desirable to some people like Media producers and Content Providers. The easy replication of media has given a boost to the illegal act of piracy, hence incurring heavy financial loss to the owner of the media.
In order to tackle this problem, a prominent solution is the use of digital watermarking. Digital watermarking consists of adding some additional information into the digital multimedia taking into consideration not to degrade the quality of the watermarked media. The additional data will help to determine the authenticity and ownership of the media.
Project Scope and Objective
Background study
Watermarking Overview
Watermarking has existed between us since the year 1982.it was first introduced in Italy. [1]Watermarking has paved its way from paper watermarking to digital watermarking.
Digital Watermarking is a process of adding/embedding a mark or label into a digital content. This embedded information is most often invisible to the human eyes or inaudible in case of audio media. Moreover, the concept of watermarking is often associated with stenography.
Stenography has been long associated with secret methods of communication. In fact, it refers to hide an important message into a cover media. Indeed, digital watermarking may be use as a mean of hiding a personal message to protect a product copyright.
However, the use of watermarking is not limited to those mentions above only. Watermarking can be also use as a method of fingerprinting, Source Tracking, authentication etc…Today watermarking has gain remarkable attention as a potential tool to fight against piracy.
Below are some watermarking examples.
Figure 1: A paper watermarking [2] Figure 2: Visible digital watermarking [3]
Purpose of watermarking
The use of watermarking is as old as paper manufacturing. In fact, watermark was primarily used to certify the composition of paper including the nature of the fiber.
Nowadays with the digitization of our world, its uses have extended to include immaterial digital impressions for use in:
Authentication
In the domain of data security, we can use digital watermark in providing certification and authentication. Certification is an important issue for official documents, such as identity cards or passport. Digital watermarking permits linking information on documents. [3]
For example, the name of a passport owner is normally printed in clear text. But it would also be hidden as an invisible watermark in the passport photo. If anyone tries to tamper with the passport by replacing the photo it would be possible to detect the change by scanning the passport and verifying the name hidden in the photo.
Source Tracking
Different recipients get differently watermarked content)
Another application is in source tracing. A watermark is embedded into a digital signal at each point of distribution. If a copy of the work is found later, then the watermark can be retrieved from the copy and the source of the distribution is known. This technique has been reportedly used to detect the source of illegally copied movies.[3]
Ownership Claim
Digital watermarking can also be adapted to mark white paper with the goal of authenticating the originator, verify the authenticity of the document content, or to date the document. Such applications are especially of interest for official documents, such as contracts. For example, the digital watermark can be used to embed the name of the lawyer or important information such as key monetary amounts. In the event of a dispute, the digital watermark is then read allowing authentication of key information in the contract.
Fingerprinting
This application of watermarking allows acquisition devices (such as video camera, audio recorders etc.) to insert authentication information about the specific device(e.g. an ID number and date of creation).furthermore this techniques is consider to be more difficult to exercise or alter the embedded info, than when using conventional digital signatures. In fact some of the modern digital cameras already include the feature.
Classification of watermarking
Watermarks can be categorized under many aspects .the figure 3 below shows a classification of watermark base on human perceptibility, inserted media, robustness and processing method.
Figure 3: classification of watermarking.
Evaluation of watermarking schemes
Source Based v/s Destination Based
Source based watermark are desirable for ownership identification or authentication where a unique watermark identifying the owner (source) is embedded to all the copies of a particular media being distributed. Source based watermarking can be used to see if a received image has been tampered with or else to check it integrity.
Whereas Destination Based Watermarking is whereby each distributed copy gets a unique watermark identifying the particular buyer. The destination based watermarking help to track the buyer in case of illegal reselling.
Spatial domain v/s Frequency domain
Watermarking of images can be done in either the spatial domain or the spectrum/frequency domain. In spatial domain many methods exist to watermark. The most used amongst them is to alter the LSB bit of the image. However if the watermarked image is subjected to any modification, the watermark recovery fails. A more robust watermark can be embedded by superimposing a symbol over an area of the picture. Still, the symbol or watermark can easily be removed by picture cropping available in any picture editing software.
To overcome all these short comes, watermarking in the frequency domain is preferred as the latter is far much better against cropping, compression and noise attack. The most common techniques use to watermark in frequency domain are DCT and DWT (to be explain further in section).
Table 1: shows the major differences associated in watermarking in spatial or frequency domain.
Visible watermarking v/s Invisible watermarking
Visible watermarks as indicated by its name itself are watermarks which can be detected by the human eyes. The stego key is embedded in the spatial domain of the cover media. This type of watermark is usually use by company to identify their content. The embedded watermark is mostly likely to be a seal or a logo. Moreover, this type of watermark can be made robust to watermarking attack by superimposing the symbol into the cover the image.
Invisible watermark, on the other hand, is not visible to the human eyes.it is embedded in the frequency domain of the cover media. Since invisible watermark are difficult to detect, they are difficult to tamper too .Moreover, Invisible watermarking is more resistant to watermark attacks such as compression and cropping.
Figure 4: an example of visible watermarking Figure 5: an example of invisible watermarking
[Google visible watermarking] [Google invisible watermarking]
Invisible fragile v/s invisible robust watermark
Fragile watermarking is use to check the integrity of the data. The stego key is embedded such a way that any tampering or modification in the cover image would cause considerable damage to the watermarked data. Hence the state of the watermark can be used to deduce whether the data has been tampered.
In the other hand robust watermark are embedded such a way that they can withstand attack such as cropping, compression, scaling and filtering. This scheme of watermarking is use for authentication purpose.
Public v/s Private watermarking
Public scheme is the far most complicated to implement as this techniques does not need the original image to recover the watermark from the watermarked image.
Private scheme make use of the original image on order to be able to decode the watermark from the stego image.
Symmetric v/s Asymmetric key watermarking
Symmetric watermarking scheme uses the same key for both the embedding and detection process. This scheme differs from the asymmetric watermarking scheme as in that, the latter uses two different keys for watermark embedding and detection.
Algorithm for a simple watermarking scheme
A watermarking process can be broken into sub process. The main sub process involves in a watermarking scheme is embedding, detection and extraction. Figure 6 below show a typical watermarking scheme.
Cover Image
Stego Image
Watermark Extractor
Watermark Detector
Watermark Embedder
Optional
Key
Watermark
Figure 6: a simple watermark algorithm.
Image watermarking techniques
Least significant bit
This technique makes use of the least significant bit of each pixel of the image to add the watermark data. The figure bellow shows an example of the famous lena picture being watermark using the least significant bit method.
Figure7: original picture Figure 8: Stego picture after LSB watermarking.
Correlation based techniques
Correlation-based digital watermarking technique for robust image pattern authentication involves hiding a phase-based signature of the image back into its Fourier magnitude spectrum in the embedding stage. The detector computes the Fourier transform of the watermarked image and extracts the embedded signature. Authentication performance is measured by a correlation test of the extracted signature and the signature computed from the watermarked image. The quality of the watermarked image is obtained from the peak signal-to-noise ratio metric.
Figure 9: Correlation based technique
DCT based.
As discussed in section the benefit of frequency domain watermarking over spatial domain is numerous. One of the classic and popular domains for image processing is the DCT.
The DCT is one of the basic building blocks of the JPEG image compression, therefore making this technique resistant to JPEG image Compression.
The process involves breaking the image into different frequency bands as shown in figure below.
Fl denotes the lowest frequency components of the block (8 x 8) while FH denotes the highest frequency components. This two frequency bands are usually discarded by compression software. Moreover, watermarking in the following frequency bands may cause severe distortion to the watermarked image.
So , fm which denotes the middle frequency components is chosen as the embedding region(shoemarker 2002) so as to provide resistance to lossy image compression ,whilst avoiding significant distortion in the watermarked image.
Figure 10: DCT watermarking
DWT watermarking techniques
The basic idea behind discrete wavelet transform (DWT) in image processing is to decompose the image into 4 independent frequency districts. The for frequency district are namely (LL) and three high frequency districts (LH,HL,HH).The low frequency district is an image close to the original image whereas the frequency districts LH,HL,HH represent the level detail, upright detail and the diagonal detail of the original image. The LL band can be further decompose using DWT as shown in figure11.
According to HVS, human eyes are more sensitive to smooth region of an image and not the tiny change of edge, profile and streak. Therefore, if we embed our watermark in the high frequency band of the dwt transformed image, we can a good concealing effect.
Figure 11.sketch map of Image DWT decomposed
Spread spectrum techniques
A Narrow-band signal is transmitted over a much larger bandwidth such that the signal energy presented in any signal frequency is undetectable. A watermark is spread over many frequency bins so that the energy in one bin is very small and certainly undetectable. Because the watermark verification process knows the location and content of the watermark, it is possible to
concentrate these weak signals into a single output with high SNR.[ http://www.cmlab.csie.ntu.edu.tw/~ipr/ipr2005/data/material/Spread%20Spectrum%20Watermarking.pdf, Prof. Ja-Ling Wu, 2005/09/26]
Watermarking Requirements
There are a number of measurable characteristics that a watermark should exhibit. These include that it should be difficult to notice, robust to common distortions of the signal, resistant to malicious attempts to remove the watermark, support a sufficient data rate commensurate with these application, allow multiple watermarks to be added and that the decoder be scalable.
Imperceptibility
Imperceptibility
The watermark should not be noticeable to the viewer, nor should then watermark degrade the quality of the original image. However, if a signal is truly imperceptible, then perceptually based lossy compression algorithms probably, still leave room for an imperceptible signal to be inserted. This may not be true for the next generation compression algorithms. Thus, to survive the next generation of lossy compression algorithms, it will probably be necessary for a watermark to be noticeable to a trained observer.
Robustness
The watermark must be difficult to remove. If only partial knowledge is available (e.g. the exact location of the watermark in an image is unknown) then attempts to remove or destroy a watermark, should result in severe degradation in fidelity before the watermark is lost.
on signal processing
Common signal processing
The watermark should still be retrievable even if common signal processing operations are applied to the data. These include digital to analog and analog to digital conversion, re-sampling and re-quantization and common signal enhancements to image contrast and color, or audio bass and treble.
Common Geometric Distortions
Watermarks in image and video data should also be immune from geometric image operations such as rotation, translation, cropping and scaling.
Watermarking constraints and attacks
Watermarking-Constraints
The success of the watermarking Scheme largely depends upon the choice of the watermark structure and insertion strategy. The two main constraints involved in the problem of watermarking are those of maintaining the robustness of the watermark information while keeping visual perception of the original image intact. If the insignificant portions of the original image are used for hiding the watermark structure then the visual perceptions of the original image may remain unaffected but the robustness of the technique decreases. On the other hand if the hiding is done in the significant portions of the original image then the robustness of the technique increases at the cost of visual perceptions. Thus the cost of function of the problem of watermarking is a weighed sum of these factors:
Cost Function, S=FR*R+FV*V,
Where,
FR = Weighted factor for Robustness,
FV = Weighted factor for Visual Perceptions,
R = Robustness,
V = Visual Perceptions.
The challenge in any watermarking technique is to maximize this cost function.
Attacks on watermarking systems
The different types of watermarking systems include common signal processing, geometric and other intentional attacks.
JPEG and MPEG Compression
The digital signals, which are discrete in, time domain results in a very large number of coefficients in the corresponding frequency domain representation. A large number of coefficients mean the necessity of more & more amount of space and increased computational time. However, it is observed that in case of audio & video signals maximum power is concentrated in the low frequency. Thus the inclusion of all the frequency coefficients results in an unnecessary increase of space and time complexity. This leads to the necessity of compression of the original source material for more efficient storage and transmission. Thus, the watermarking strategy should be such that the watermark structure is not lost or distorted beyond recognition while discarding the high frequency coefficients during compression.
Filtering
The marketed image may undergo several filtering operations in the course of enhancement techniques. Such filtering operations include low pass, high pass, median or Gaussian filtering. It is absolutely necessary for the watermarked information to be resistant to all such filtering processes.
Rescaling
The watermark is also prone to be distorted or lost while the marketed image is rescaled.
Cropping
A common editing operation is the spatial rejection of a portion of the marketed image. Such an operation is also considered to be a serious threat to the watermarked information. This is a very important consideration when the watermarking is done in the spatial domain. This requires the watermarked information to be embedded in those areas of the original image which are significant, and if cropped away results in severe degradation of the actual image.
Alternative to watermarking
Copyright your Photo
One of the simplest alternative to watermarking is to copyright your photo. In fact, in some countries such as US, UK and Canada; the photo is automatically copyrighted as from it creation. Copyright your photo enables you to have all the necessary rights to recover any lost fees from unauthorized use.
Hidden Layers.
Place the actual image behind a transparent foreground image. The online image will appear completely normal, but whenever someone tries to right click and save it, the resulting file will be the blank foreground image â€" not the background.
Tiling.
 Another option is to break your photo into smaller image tiles. These will appear as a continuous image, but anyone trying to copy your image will end up saving only one tile piece at a time. However, the tiling approach also means that your web server will have to provide several files instead of just one â€" which may make the web page load more slowly, and can potentially increase your server requirements. Separating each photo into smaller tiles can also be quite time consuming, so this option isn't commonly used.
Disable right-clicks.
This method can be a very effective way of letting your visitors know you are serious about image protection, and of ensuring that they receive a copyright notice whenever they try to save your image
