International Core Journal of Engineering 2020-26 | Page 143

Find the Roberts gradient for the pixel point f ( i , j ) :
' x 2 f  ' y 2 f
R( i , j )
In the formula,
(3)
3) Select an action
The experimental vector u i g  1 generated after the
variation is crossed is compared with the parent individual
x i g . If the experimental vector fitness function is better than
Therefore, the gradient value for the original image
f ( x , y ) can be approximated as:
R ( x , y )
^ > f ( x , y )  f ( x  1, y +1) @ + > f ( x  1, y )  f ( x , y  1) @ `
1
2 2
x i g , then select u i g  1 Enter the next generation, otherwise
(4)
choose x i g as the child, choose the formula as follows:
After calculating the gradient value through the above
steps, the appropriate threshold T is selected. When R ( x , y )
is greater than or equal to T, it is an edge point, since most of
the information of the image exists and the edge of the
copyright image Part, therefore, indicates that this point
contains the main information of the image. In this paper, the
Roberts edge detection operator is used to preprocess the
copyright image, which not only reduces the changes to the
database value, but also embeds the main information of the
image into the database. After the watermark is extracted, the
copyright attribution can be clearly determined.
x i g  1
­ u i g  1
® g
¯ x i
f ( u i g  1 )  f ( x i g )
f ( u i g  1 ) t f ( x i g )
(7)
Where f represents the individual fitness function.
III. R EVERSIBLE DATABASE WATERMARKING ALGORITHM
BASED ON DE ALGORITHM
This section introduces the proposed differential
evolution reversible database watermarking algorithm. The
reversible database watermark mainly includes three parts. (1)
watermark preprocessing; (2) watermark embedding; (3)
watermark extraction and data recovery. The preprocessing
stage is divided into copyright image preprocessing and
database preprocessing. The watermark embedding stage
embeds the copyright image into the best embedding position
obtained by the differential evolution algorithm to ensure the
invisibility and usability of the watermark database. After the
watermark is embedded, the database is robust and invisible,
and the watermark extraction and data recovery are
embedded in the inverse process.
B. Differential evolution algorithm
The differential evolution algorithm DE (Differential
Evolution Algorithm) is a heuristic random search algorithm
based on group difference. It was proposed by R. Storn and
K. Price in 1997 to solve the Chebyshev polynomial. The
algorithm has simple principle and controlled parameters.
Less, strong robustness, etc. The algorithm includes the
following three basic operations:
1) Mutation operation
First, N D-dimensional populations are initialized. In the
g-generation iteration, three individuals x r 1 , x r 2 , and
Finally, complete content and organizational editing
before formatting. Please take note of the following items
when proofreading spelling and grammar:
g
x r 3 different from the target vector x j are randomly selected
from the population, so the population The scale NŎ4, the
variation vector generated by the mutation formula is:
v i g  1
rand j  (0,1), sn is a random number
in {1, 2,..., D}, D is the individual dimension, and j is the j激
th variable of the i-th individual.
The algorithm flow chart is shown in Figure 2:
x r 1 g  F *( x r 2 g  x r 3 g ) (5)
Where F is the scaling factor, i is the i-th individual, and
F is too small to cause the algorithm to be premature. If the
value is too large, the algorithm does not easily converge to
the optimal value, generally between [0, 1].
2) Cross operation
Cross-operation is a necessary step to improve the
diversity of the population. For the g-generation population,
the target vector individual x i g and the mutated individual
v i g  1 are subjected to partial gene cross-operation according
to the following formula. Generate experimental vector
u i , j g  1 漡
g  1
i , j
u
­ ° v i g ,  j 1 if ( rand j [0,1] d CR ) or ( j
® g
otherwise
°̄ x i , j
sn )
Fig 2. Algorithm flow
A. Watermark preprocessing stage
Firstly, the copyright image is preprocessed by the
above-mentioned Roberts edge detection operator. After
preprocessing, a pair of black and white binary images are
obtained, and the binary image obtained by the edge
(6)
121