Research on image steganography based on conditional Invertible Neural Network
Abstract
To
improve
the
security
of
image
steganography,
an
steganography
method
based
on
conditional
Invertible
Neural
Network
is
proposed
in
this
paper.
First,
we
design
a
to
obtain
high-quality
stego
images
with
rich
high-level
semantic
information
and
clear
spatial
details.
Based
directivity
Network,
can
accurately
adjust
ensure
controllability
content.
We
introduce
dual
cross-attention
module
into
network
structure.
The
integration
modules
enhances
feature
extraction
captures
complex
details
steganographic
accuracy.
In
addition,
introduction
convolutional
block
attention
layer
direct
model's
focus
key
regions,
refining
quality.
increase
number
blocks,
which
improves
efficiency
reuse.
A
large
experiments
are
carried
out
datasets.
For
cover
pairs,
PSNR
value
reached
43.62dB,
for
secret
recovery
46.48dB.
Experimental
results
show
that
quality
better
than
other
state-of-the-art
methods.
Research Square (Research Square), Год журнала: 2024, Номер unknown
Опубликована: Дек. 13, 2024
Язык: Английский