Digital Signal Processing Reference
In-Depth Information
Fig. 4.28
Basic idea of the two-description image coding with steganography
in the spatial domain is simple, while the block partition in the wavelet domain
normally produces better coding performance, where the wavelet coefficients in
different sub-bands corresponding to the same spatial location are grouped to form
tree-structured blocks. Figure
4.28
shows such a block checkerboard partition in the
wavelet domain.
4.4.2.1
Fine and Coarse Coding
Two sub-images are grouped by part 1 and part 2 blocks in Fig.
4.28
, respectively.
The wavelet coefficients of part 1 sub-image are quantized by a coarse quantizer,
and the quantized coefficients are referred to be coarse information. The wavelet
coefficients of part 2 are quantized by a fine quantizer to produce fine information.
The coarse information is embedded (hidden) into the fine information. Note that
the length of the coarse information (denoted as L) needs to be transmitted for
extracting purpose, and the bit cost for the length is trivial and negligible. For
example, for a 512
512 image with 0.1 bpp coarse information, the bit cost for
the length representation is about log
2
0:1
512
2
D
15 bits, which amounts to an
overhead of 15=.512
512/
D
5:9
10
5
bit-per-pixel (bpp).
4.4.2.2
Embedding Approach - LSB Steganographic Method
Steganography can send message under the cover of a carrier signal, and many
techniques of steganography have been proposed [
33
,
35
]. One of the well-known
steganographic methods is the LSB substitution which replaces the least-significant
bits of coefficients with the information bits. In this section, the LSB steganographic
method is exploited to embed the coarse information in the least-significant bits of
the fine information selectively. The process of the LSB steganographic method is
shown in Fig.
4.29
.
The principle of the LSB steganographic method is to change the LSB of nonzero
quantized wavelet coefficients to match the coarse information bit stream in the
following way:
