Other Free Encyclopedias » Online Encyclopedia » Encyclopedia - Featured Articles » Contributed Topics from F-J

Image Watermarking - Challenges and benchmarking, Advanced, Region of interest watermarking, Self-correcting images

algorithms changes information algorithm

Definition: Image watermarking deals with creating a metadata (a watermark) about the image content and hiding it within the image.

Challenges and benchmarking

Digital images are often printed and scanned again, for example when they are used in magazines and readers want to store digitally. When used as web site illustrations, they are often compressed by lossy compression algorithms like JPEG. Common processing operations include softening, sharpening, denoising, scaling, cropping and colour corrections. A number of benchmarking suits address image watermarking robustness evaluation, Stirmark, Checkmark and Certimark are well known examples.

The most challenging attacks on the robustness of image watermarks today are nonlinear transformations, rendering a watermark in an image undetectable for the watermarking algorithm. But even common image processing operations like scaling and rotation can be a serious challenge for the re-synchronisation of an image watermark.

Typical challenges in practical solutions are printing and scanning with low quality devices like customer ink jet printers leading to the additions of noise, colour changes, small rotations and changes in image resolution.

Another important aspect of image watermarking is the broad variety of image types. There are photos, also called natural images in the literature, figures based on line drawings, rendered synthetic images, bitmap representations of textual information, just to give the most common examples. Challenges with respect to transparency and robustness often depend on the characteristics of these images. Many watermarking algorithms address only one image type, like e.g. photographs. Often the dependency can be derived from the embedding principle: An algorithm which needs textures to embed information will not be able to deal with a black and white drawing but performs perfectly with most photographs.

Advanced image watermarking

While all watermarking methods described in this article are known for image watermarking, there are also a number of innovative approaches which are especially suited for image watermarking. They usually could be transferred to video watermarking, but the long computation time caused by high complexity and the stronger compression in digital video are a serious hindrance here. As examples for advanced watermarking methods we describe two innovative algorithms combining existing approaches with new techniques.

Region of interest watermarking

A good example of advanced fragile watermarking for integrity protection is region of interest (ROI) watermarking. The basic idea is distinguish between semantically important and unimportant regions of an image. It is often sufficient to protect the important regions as manipulations changing the meaning of the image will only occur in these. Modifications taking place not in the RIO may be of cosmetic nature or could be simple name tag annotations. An example of ROI watermarking is the face detection approach discussed in protecting only automatically detected faces and the relative positions in an image.

Self-correcting images

While many approaches to identify changes in images are known, some algorithms even allow re-creating the original from the watermark to a certain degree. This enables a comparison of the content of an image before and after an attack. One known approach is based on an algorithm described in, using a block-based pseudo-random distribution of pixel information in over the image thereby shuffling and spreading a kind of parity checksum over the whole image. The idea here is to use the spread information about a block of pixels to identify small local changes by identifying parity errors after attacks.

Image Watermarking Using Visual Cryptography [next] [back] Image Secret Sharing - A (k,n)-threshold scheme, Basis matrices in visual cryptography

User Comments

Your email address will be altered so spam harvesting bots can't read it easily.
Hide my email completely instead?

Cancel or