The Imaging Science Journal,
Год журнала:
2022,
Номер
70(2), С. 75 - 86
Опубликована: Фев. 17, 2022
Segmentation
of
wound
images
is
important
for
efficient
treatment
so
that
appropriate
methods
can
be
recommended
quickly.
Wound
measurement,
subjective
an
overall
assessment.
The
establishment
a
high-performance
automatic
segmentation
system
great
importance
care.
use
machine
learning
will
make
performing
with
high
performance
possible.
Great
success
achieved
deep
learning,
which
sub-branch
and
has
been
used
in
the
analysis
recently
(classification,
segmentation,
etc.).
In
this
study,
pressure
was
discussed
different
encoder-decoder
based
models.
All
are
implemented
on
Medetec
image
dataset.
experiments,
FCN,
PSP,
UNet,
SegNet
DeepLabV3
architectures
were
five-fold
cross-validation.
Performances
models
measured
experiments
it
demonstrated
most
successful
architecture
MobileNet-UNet
99.67%
accuracy.
Scientific Programming,
Год журнала:
2022,
Номер
2022, С. 1 - 8
Опубликована: Июль 31, 2022
Medical
image
segmentation
identifies
an
area
that
should
be
analyzed
later
in
the
processing
process,
such
as
for
disease
recognition
and
classification.
As
search
is
reduced,
this
action
allows
faster
computation
analysis.
We
propose
use
of
a
heuristic
red
fox
optimization
algorithm
(RFOA)
medical
paper.
The
heuristics’
operation
was
adapted
to
analysis
two-dimensional
images,
with
focus
on
equation
modification
novel
fitness
function.
proposed
solution
analyzes
by
converting
selected
pixels
one
two
color
variants,
black
or
white,
based
threshold
value
used.
Their
number
counted,
allowing
chosen
threshold.
result,
results
automatic
selection
parameter.
Our
method
new
function
adjustment
RFOA
used
publicly
available
database
lung
X-ray
images
evaluation,
results,
accuracy
performed,
well
discussion
benefits
drawbacks
presented.
Journal of Cloud Computing Advances Systems and Applications,
Год журнала:
2023,
Номер
12(1)
Опубликована: Март 21, 2023
Abstract
The
worldwide
usage
of
Internet
Things
(IoT)
applications
enhanced
the
utilization
consumer
devices,
such
as
smartphones,
computers,
screening
equipment
used
in
hospitals
that
merely
rely
on
imaging
techniques.
Numerous
images
got
generated
over
cloud
platform
a
daily
basis
ad
create
storage
complexity.
On
other
hand,
securing
data
stored
is
important.
Instead
storing
large
amount
into
cloud,
lightweight
dynamic
processing
suppresses
complex
issues
security.
Here
secure
cloud-based
image
architecture
discussed.
Privacy
preserving
medical
communication
considered
specific
research
scope.
Cryptographic
technique
to
encode
original
and
decode
at
end
currently
conventional
design.
Providing
privacy
records
through
adding
noise
denoising
same
proposed
idea.
work
keenly
focused
creating
light
weight
communicates
effectively
with
perseverance
using
deep
learning
technique.
In
system,
design
an
efficient
scheme
hybrid
classification
model
created
ensure
reliable
communication.
Deep
algorithms
merged
form
Pseudo-Predictive
Denoising
Network
(PPDD).
system's
benefit
ensuring
added
security
Dark
Cloud
newly
structured
algorithm.
packed
Gaussian
act
key.
complete
packing
unpacking
encapsulated
by
transformed
images.
Over
premise,
highly
secured
invisible
malicious
users.
To
reduce
complexity,
unpacked
denoise
process
applied
edge
devices.
During
authorized
access
period
alone,
decrypted
accessible
nodes.
maximum
dynamically
happen
without
depending
boundary.
performance
PPDD
network
evaluated
Signal
ratio
(SNR),
Similarity
index
(SI),Error
Rate(ER)
Contrast
ratio(CNR).
comparatively
validated
existing
state-of-art
approach.
Heliyon,
Год журнала:
2023,
Номер
9(4), С. e15097 - e15097
Опубликована: Апрель 1, 2023
As
an
important
step
in
image
processing,
segmentation
can
be
used
to
determine
the
accuracy
of
object
counts,
and
area
contour
data.
In
addition,
is
indispensable
seed
testing
research.
Due
uneven
grey
level
original
image,
traditional
watershed
algorithms
generate
many
incorrect
edges,
resulting
oversegmentation
undersegmentation,
which
affects
obtaining
phenotype
information.
The
DMR-watershed
algorithm,
improved
algorithm
based
on
distance
map
reconstruction,
proposed
this
paper.
According
distribution
characteristics
reduction
amplitude
h
was
selected
mask
with
same
trend
as
that
image.
greyscale
reconstructed
corresponding
thresholds
according
false
minima
different
regions
are
segmented,
generates
accurate
eliminates
wrong
edges.
An
adzuki
bean
(Vigna
angularis
L.)
experimental
material
residual
rate
counting
results
each
investigated
two
cases
two-particle
adhesion
multiparticle
adhesion.
were
compared
those
edge
detection
concave
point
analysis
commonly
for
segmentation.
case
adhesion,
rates
0.233
0.275,
respectively,
while
0
proved
suitable
not
applicable
because
it
would
destroy
0.063
0.188,
reached
100%,
effectiveness
algorithm.
paper
significantly
improves
adherent
seeds,
provides
a
new
reference
processing
testing.
