Cryptography Reference
In-Depth Information
Fig. 13.6.
(Top) Embedding phase and (bottom) extraction phase of the proposed
lossless data-embedding algorithm.
In the extraction phase, the watermarked signal s
w
is quantized and the
watermark payload h which is the compressed residual and the payload data is
extracted in Eq. (13.9). The residual r is decompressed by Q
L
(s
w
)=s−Q
L
(s)
as side information. The original host is reconstructed by replacing the lowest
levels of the watermarked signal by the residual in Eq. (13.10):
r
=
s−Q
L
(s),
(13.9)
s
=
Q
L
(s)+r = Q
L
(s
w
)+r.
(13.10)
The lossless-embedding capacity of the system is given by
C
Lossless
= C
GLSB
−C
residual
,
(13.11)
where C
Lossless
is the raw capacity of GLSB embedding (C
GLSB
= log
2
(L))
and C
residual
is the capacity consumed by the compressed residual. To further
improving the lossless embedding capacity, Celik et al. adopt CALIC lossless
image compression algorithm [13, 15]. This uses the unaltered portions of the
host signal, Q
L
(s) as side-information, to e
ciently compress the residual.
From what is reported in [8], Celik et al. applied several test images, F-
16, Mandrill and Barbara to the lossless G-LSB algorithm with its selective
embedding extension, Celik et al. compare the results with the RS embed-
ding scheme [7]. The amplitude of the flipping function varied from 1 to 6.
The lossless G-LSB algorithm at 100% embedding outperforms the RS em-
bedding scheme from a capacity distortion perspective at most points except
for the lowest distortion points at A =1andL = 2. The reason is that
RS embedding modifies the pixels corresponding to the R and S groups while
skipping U groups. By modifying the embedding extension from 100% to 75%,
the lossless G-LSB algorithm slightly surpasses RS embedding at the lowest