Information Technology Reference
In-Depth Information
select
DCT
Selected DCT
coefficients
Image in space do-
main: blocking
Image in block
DCT domain
shuffle
dither
DFT
Dither modulated
DFT coefficients
DFT of shuffled
DCT coefficients
Shuffled DCT
coefficients
Fig. 6. Double transform/QIM data embedding
From (9), the presence of embedded data in the second-layer transform coefficients
can be detected by analyzing the energy spectrum of the first-layer coefficients. This
is illustrated in Fig. 7, which shows the square root of energy, P , in the block-DCT
coefficients of image Lena before and after data embedding in the 30th transform-
domain position. The abscissa represents zigzag-sorted frequency index N of the 8×8
block-DCT. The peak at N =30 in Fig. 7(b) is a clear signature of data embedding.
Height of the peak is related to the embedding strength, i.e., the quantization step ∆.
50
50
40
40
30
30
P
P
20
20
10
10
0
0
0
20
40
60
0
20
40
60
N
N
Fig. 7. Spectrum of the square root of energy in the DCT coefficients, P , with respect to zigzag-
ordered DCT index, N , for the original and watermarked images
5
Experimental Results
In the experiment, watermarking schemes based on block transform and QIM, both
single-layered and double-layered, were used to embed binary sequences into test
images Lena, Baboon and Cameraman, as well as a set of black-and-white images in a
database generated by color-to-grayscale conversion from pictures taken with a digital
Search WWH ::




Custom Search