Cryptography Reference
In-Depth Information
Fig. 13.11.
Embedding process of the improved DE method.
According to a predefined scanning order, all the difference values are
ordered as a one dimensional listh
1
,h
2
,...,h
n
. Then, difference values are
classified into four classes according to Fig. 13.12. A location map denoted as
lM ap is created to record the positions of Class 1, Class 2 and the Flipping
Class 2. For the h in Class 1, we assign a value 1 in the lM ap; for the value h
in Class 2, we assign a value 0 in the lM ap. lM ap and pM ap are concatenated
to form a sub-bitstream l. l is losslessly compressed using arithmetic coding.
The compressed sub-bitstream is denoted as the L.
In Class 1 h is kept intact. Considering h in the Flipping Class 2. If the
embedded bit is 1, then h
′
f
via Eq. (13.14) is equal to 2h + 1, otherwise
the value of h
′′
f
is 2h. We select h
′
f
if h
′
f
does not satisfy the the condition
Eq. (13.15), otherwise h
f
is selected. This ensures that the difference values
in the Flipping Class 2 will be classified into the Class 2 during extraction.
All of L is embedded in the LSB space saved by the bit-shift operation over
h of Class 3. Then most of real watermark bits including the hash function of
image block, P , are embedded in the space saved by the bit-shift operation
over the remanent h in Class 3. The remaining values of P are embedded
into LSBs of the one-bit left shifted versions of h in Class 2. That is with the
exception of Flipping Class 2.
′′
Watermarking Decoding Scheme Based on the Improved
Difference Expansion Method
The general framework of the watermark detection scheme is shown in
Fig. 13.13. The locating decoding pattern is created by tiling with the basic
