Choosing A Suitable Digital Watermark
[ Home Page - Surveys - Documents - Presentations - Search ]
Watermarking is an effective technology that solves many problems within a digitisation project. By embedding Intellectual Property data (e.g. the creator, licence model, creation date or other copyright information) within the digital object, the digitiser can demonstrate they are the creator and disseminate this information with every copy, even when the digital object has been uploaded to a third party site. It can also be used to determine if a work has been tampered with or copied.
This paper describes methods for establishing if a project requires watermarking techniques and criteria for choosing the most suitable type.
Before implementing watermarking within your workflow, you should consider its proposed purpose. Are you creating watermarks to indicate your copyright, using it as a method of authentication to establish if the content has been modified, or doing so because everyone else has a watermarking policy? The creation of a watermark requires significant thought and modification to the project workflow that may be unnecessary if you do not have a specific reason for implementing it.
For most projects, digital watermarks are an effective method of identifying the copyright holder. Identification of copyright is encouraged, particularly when the work makes a significant contribution to the field. However, the capabilities of watermarks should not be overstated. It is useful in identifying copyright, but is incapable of preventing use of copyrighted works. The watermark may be ignored or, given sufficient time and effort, removed entirely from the image. If the intent is to restrict content reuse, a watermark may not be the most effective strategy.
To assist the choice of a watermark, the project team should identify the required attributes of a watermark by answering two questions:
The answer to the first question is influenced by the skills and requirements of your target audience. If the copyright information is intended for non-technical and technical users, a visible watermark is the most appropriate. However, if the copyright information is intended for technical users only or the target audience is critical of visible watermarks (e.g. artist may criticise the watermark for impairing the original image), an invisible watermark may be the best option.
To answer the second question, the project team should consider the purpose of the watermark. If the intent is to use it as an authentication method (i.e. establish if any attempt to modify the content has been made), a fragile watermark will be a valued attribute. A fragile watermark is less robust towards modifications where even small change of the content will destroy embedded information. In contrast, if the aim is to reflect the owner's copyright, a more robust watermark may be preferential. This will ensure that copyright information is not lost if an image is altered (through cropping, skewing, warp rotation, or smoothing of an image).
If resilience is a required attribute of a digital watermark, the project team has two options: invisible or visible watermark. Each option has different considerations that make it suitable for specific purposes.
Invisible Watermarks
Invisible watermarks operate by embedding copyright information within the image
itself. As a rule, watermarks that are less visible are weaker and easier to
remove. When choosing a variant it is important to consider the interaction between
watermark invisibility and resilience. Some examples are shown in Table 1:
| Name | Description | Resilience |
| Bit-wise | Makes minor alterations to the spatial relation of an image | Weak |
| Noise Insertion | Embed watermark within image noise | Weak |
| Masking and filtering | Similar to paper watermarks on a bank note, it provides a subtle, though recognisable evidence of a watermark. | Strong |
| Transform domain | Uses dithering, luminance, or lossy techniques (similar to JPEG compression) on the entire or section of an image. | Strong |
Table 1: Indication of resilience for invisible watermarks
'Bit-wise' & 'noise insertion' may be desirable if the purpose is to determine whether the medium has been altered. In contrast, 'transform domain' and 'masking' techniques are highly integrated into the image and therefore more robust to deliberate or accidental removal (caused by compression, cropping, and image processing techniques) in which significant bits are changed. However, these are often noticeable to the naked eye.
Visible Watermarks
A visible watermark is more resilient and may be used to immediately identify
copyright without significant effort by the user. However, these are, by design,
more intrusive to the media. When creating a visible watermark, the project team
should consider its placement. Projects funded with public money should be
particularly conscious that the copyright notice does not interfere with the
purpose of the project. A balance should be reached between the need to make the
watermark difficult to remove and its use to the user.
Both watermarks make them suitable for specific situations. If handling a small image collection, it may be feasible (in terms of time and effort) to use both watermarks as a redundant protection measure - in the event that one is removed, the second is likely to remain.
If the project is using a watermark to establish their copyright, some thought should be made on the static information you wish to provide. For example:
Some content management systems are also able to generate dynamic watermarks and embed them within the image. This may record the file information (file format, image dimensions, etc.) and details about the download transaction (transaction identifier, download date, etc.). This may be useful for tracking usage, but may annoy the user if the data is visible.
To avoid unnecessary corruption of a watermark by the digitiser/creator themselves, the watermark creation process should be delayed until the final steps of the digitisation workflow. Watermarks can be easily removed when the digitiser is modifying the image in any way (e.g. through cropping, skewing, adjustment of the RGB settings, or through use of lossy compression). If an image is processed to the degree that the watermark cannot be recognized, then reconstruction of the image properties may be possible through the use of an original image.
[ Home Page - Surveys - Documents - Presentations - Search ]
Web page by Brian Kelly of UKOLN.
Last Modified 20-April-2006
Email comments to webmaster@ukoln.ac.uk