It provides capacity bounds for reversible data hiding based on the HVS. In addition, in order to avoid truncation at gray scale limits each modulated value must reside within a certain range. The hash h2 ensures the integrity of visible mark Y. Finally, thanks to NUS for its financial assistance without which it would have been impossible for me to continue study in abroad. The second category is further divided into spatial domain lossless watermarking and frequency domain lossless watermarking algorithms.

We extract the whole watermark using this side information and decompress the extracted information as necessary. The authentication information, say the hash of the original image, and other necessary information are concatenated with the compressed bit- string. Note that, the results given in Table 2. Moreover, the amount of side information is very low and so the embedded capacity is not consumed greatly for passing side information to the decoder and is free for embedding large amount of necessary information like electronic patient records EPR in medical images. Nevertheless, even a tiny change in the original x-ray image by intentional or unintentional attackers could lead to take erroneous medical decisions. Kankanhalli, Rajmohan, and K. Satellite image of northeast Pacific a original image with Hawaii islands, b cropped image without Hawaii islands 1.

This type of attack is just the complement of the statistical averaging attack. Another possible attack is performed on the hardware components of the watermarking system. Finally, we develop a new variation of the use of reversible watermarking — incremental watermarking.

In the following subsections, we discuss this type of algorithms briefly. In contrast, in the cryptographic technique authentication information like hash has to be appended as extra metadata. The images are taken from very large distances.


reversible watermarking thesis

The authors suggest using JBIG lossless compression algorithm [41]. Disadvantages are — i it is not suitable for images of smaller sizes, ii a lot of the embedding rsversible is used up by digital signatures of blocks, iii artifacts become visible when embedding level increases, iv it increases the file size, and v the capacity is low.

reversible watermarking thesis

watermarkimg The authors suggest using JBIG lossless compression algorithm [41]. For example, a slight change in the medical image of a patient could make doctors taking wrong decisions. Nevertheless, very high capacity reversible watermarking algorithm must be involved for lossless compression and watermarking of videos. They also propose to use overlapping groups instead of disjoint groups and multiple passes instead of single pass along the groups while embedding message.

Consequently, this reversoble offers low capacity.

thdsis This leaves only a few or no pixels of textured blocks without embedding data. RAC codes 76 Table 5. Take the quantized DCT coefficients, and r 4. Calculate the hash of retrieved image and compare it with extracted hash, h1.

(PDF) Reversible Watermarking using the Human Visual system | Mohammad Awrangjeb –

Embedding means adding, casting or etching the watermark data to the cover work. Pixels at textured blocks do not offer themselves for high compression ratio. Skip to main content. The second algorithm works much better on the textured images than other existing algorithms do.

We use arithmetic coding for compressing the extracted features as well as first side information and run length coding, for example relative address coding RAC briefly described in section 3. Unsuccessful extracted watermarked image is passed through affine correction and weighted extraction phases, and finally if it still fails to extract data successfully the watermarked image is declared as inauthentic. Relative address coding RAC 75 Figure 4.


That is why, the compressed data, passed as a part of the watermark, may be very large. The symbol a2, for example, is associated with subinterval [0. Secondly, the embedded data becomes an inherent part of the content, and is robust against lossless file format conversions.

It is observed that DCT coefficients with higher frequencies have a large number of zeros or very small integers 1, -1 etc. MD5 is vulnerable to cryptanalysis whereas SHA-1 appears not to be vulnerable to such attacks [45]. After finishing processing we get back the original image. The authors suggest using JBIG lossless compression algorithm [41]. Consequently, to devise a reversible watermarking algorithm offering high embedding capacity with acceptable visual quality has become an active research area.

In the second algorithm, we extract 4 to 5 LSBs from each pixel of textured blocks only and arrange them in a bit-string. Any type of simple change could give inappropriate knowledge to researchers. We will limit the use of this term in our thesis in order to avoid chaos and will use any other synonyms throughout the thesis for the same meaning.

This simple method offers high capacity with better visual quality. In most of the cases, such information is added to the authentication information before embedding and high capacity watermarking algorithms are required.