Sensors,
Journal Year:
2025,
Volume and Issue:
25(5), P. 1482 - 1482
Published: Feb. 28, 2025
Image
distortion
correction
is
a
fundamental
yet
challenging
task
in
image
restoration,
especially
scenarios
with
complex
distortions
and
fine
details.
Existing
methods
often
rely
on
fixed-scale
feature
extraction,
which
struggles
to
capture
multi-scale
distortions.
This
limitation
results
difficulties
achieving
balance
between
global
structural
consistency
local
detail
preservation
distorted
images
varying
levels
of
complexity,
resulting
suboptimal
restoration
quality
for
highly
To
address
these
challenges,
this
paper
proposes
dynamic
channel
attention
network
(DCAN)
correction.
Firstly,
DCAN
employs
design
utilizes
the
optical
flow
effectively
balancing
under
distortion.
Secondly,
we
present
fusion
selective
module
(CAFSM),
dynamically
recalibrates
importance
across
By
embedding
CAFSM
into
upsampling
stage,
enhances
its
ability
refine
features
while
preserving
integrity.
Moreover,
further
improve
consistency,
comprehensive
loss
function
designed,
incorporating
similarity
(SSIM
Loss)
optimization.
Experimental
widely
used
Places2
dataset
demonstrate
that
achieves
state-of-the-art
performance,
an
average
improvement
1.55
dB
PSNR
0.06
SSIM
compared
existing
methods.
2021 IEEE/CVF International Conference on Computer Vision (ICCV),
Journal Year:
2023,
Volume and Issue:
unknown, P. 8048 - 8059
Published: Oct. 1, 2023
Multi-modality
image
fusion
aims
to
combine
different
modalities
produce
fused
images
that
retain
the
complementary
features
of
each
modality,
such
as
functional
highlights
and
texture
details.
To
leverage
strong
generative
priors
address
challenges
unstable
training
lack
interpretability
for
GAN-based
methods,
we
propose
a
novel
algorithm
based
on
denoising
diffusion
probabilistic
model
(DDPM).
The
task
is
formulated
conditional
generation
problem
under
DDPM
sampling
framework,
which
further
divided
into
an
unconditional
subproblem
maximum
likelihood
subproblem.
latter
modeled
in
hierarchical
Bayesian
manner
with
latent
variables
inferred
by
expectation-maximization
(EM)
algorithm.
By
integrating
inference
solution
iteration,
our
method
can
generate
high-quality
natural
cross-modality
information
from
source
images.
Note
all
required
pre-trained
model,
no
fine-tuning
needed.
Our
extensive
experiments
indicate
approach
yields
promising
results
infrared-visible
medical
fusion.
code
available
at
https://github.com/Zhaozixiang1228/MMIF-DDFM.
Journal of Visual Communication and Image Representation,
Journal Year:
2024,
Volume and Issue:
101, P. 104179 - 104179
Published: May 1, 2024
Infrared
and
visible
image
fusion
represents
a
significant
segment
within
the
domain.
The
recent
surge
in
processing
hardware
advancements,
including
GPUs,
TPUs,
cloud
computing
platforms,
has
facilitated
of
extensive
datasets
from
multiple
sensors.
Given
remarkable
proficiency
neural
networks
feature
extraction
fusion,
their
application
infrared
emerged
as
prominent
research
area
years.
This
article
begins
by
providing
an
overview
current
mainstream
algorithms
for
based
on
networks,
detailing
principles
various
algorithms,
representative
works,
respective
advantages
disadvantages.
Subsequently,
it
introduces
domain-relevant
datasets,
evaluation
metrics,
some
typical
scenarios.
Finally,
conducts
qualitative
quantitative
evaluations
results
state-of-the-art
offers
future
prospects
experimental
results.
