Abstract
In this work, an enhanced watermarking method is proposed. The
SVD transform is utilized to compress the watermark image, so the
capacity, imperceptibility, and security are increased. The cover
image is separated into blocks, and the embedding is carried out on
the highest variance blocks to increase the imperceptibility of the
method. The selected blocks are transformed using the Discrete
Cosine Transform (DCT), a matrix including the DC coefficients is
created, and transformed by the Singular Value Decomposition
(SVD). The compressed watermark’s principal components are
embedded into the singular values of the cover image. A robust
method is obtained by using the DCT-SVD transforms. A watermark
with different sizes can be embedded without association with the
number of image blocks, while the size of the watermark in a similar
method is restricted by the number of cover image blocks in rows
and columns. High Peak Signal to Noise Ratio (PSNR) values are
demonstrated by experimental results lying in the range (45.5-48.5)
for the watermarked images of size 512*512 and a watermark of size
64*64, and it is possible to extract an acceptable quality watermark
from the attacked image.
SVD transform is utilized to compress the watermark image, so the
capacity, imperceptibility, and security are increased. The cover
image is separated into blocks, and the embedding is carried out on
the highest variance blocks to increase the imperceptibility of the
method. The selected blocks are transformed using the Discrete
Cosine Transform (DCT), a matrix including the DC coefficients is
created, and transformed by the Singular Value Decomposition
(SVD). The compressed watermark’s principal components are
embedded into the singular values of the cover image. A robust
method is obtained by using the DCT-SVD transforms. A watermark
with different sizes can be embedded without association with the
number of image blocks, while the size of the watermark in a similar
method is restricted by the number of cover image blocks in rows
and columns. High Peak Signal to Noise Ratio (PSNR) values are
demonstrated by experimental results lying in the range (45.5-48.5)
for the watermarked images of size 512*512 and a watermark of size
64*64, and it is possible to extract an acceptable quality watermark
from the attacked image.
Keywords
DCT transform.
Image compression
image watermarking
SVD transform