2021 2nd International Conference on Smart Electronics and Communication (ICOSEC),
Journal Year:
2021,
Volume and Issue:
unknown, P. 918 - 921
Published: Oct. 7, 2021
Games
have
evolved
from
the
beginning
of
human
civilization
to
modern
times.
One
oldest
and
most
popular
are
role-playing
games,
which
allow
player
enjoy
playing
a
role
in
game
play
variety
characters.
In
this
paper,
similarity
redundancy
3D
object
information
used,
CS
technology
is
combined
improve
computational
efficiency,
while
alleviating
high
energy
consumption
sampling
transmission
process
diffraction
field.
whole
experiment,
paper
uses
visualization
tools
analyze
summarize
data
experiment.
The Visual Computer,
Journal Year:
2022,
Volume and Issue:
39(10), P. 4555 - 4571
Published: July 25, 2022
Abstract
Depth
information
is
useful
in
many
image
processing
applications.
However,
since
taking
a
picture
process
of
projection
3D
scene
onto
2D
imaging
sensor,
the
depth
embedded
image.
Extracting
from
challenging
task.
A
guiding
principle
that
level
blurriness
due
to
defocus
related
distance
between
object
and
focal
plane.
Based
on
this
widely
used
assumption
Gaussian
blur
good
model
for
blur,
we
formulate
problem
estimating
spatially
varying
as
classification
problem.
We
solve
by
training
deep
neural
network
classify
patches
into
one
20
levels
blurriness.
have
created
dataset
more
than
500,000
size
$$32\times
32$$
32×32
which
does
not
require
human
labelling.
The
train
test
several
well-known
models.
find
MobileNetV2
suitable
application
its
low
memory
requirement
high
accuracy.
trained
determine
patch
then
refined
applying
an
iterative
weighted
guided
filter.
result
map
carries
degree
each
pixel.
compare
proposed
method
with
state-of-the-art
techniques
demonstrate
successful
applications
adaptive
enhancement
magnification
limited
images
present
clear
distinction
levels.
PLoS ONE,
Journal Year:
2024,
Volume and Issue:
19(7), P. e0303633 - e0303633
Published: July 9, 2024
Estimating
the
densities
of
marine
prey
observed
in
animal-borne
video
loggers
when
encountered
by
foraging
predators
represents
an
important
challenge
for
understanding
predator-prey
interactions
environment.
We
used
images
collected
during
trip
one
chinstrap
penguin
(
Pygoscelis
antarcticus
)
from
Cape
Shirreff,
Livingston
Island,
Antarctica
to
develop
a
novel
approach
estimating
density
Antarctic
krill
Euphausia
superba
activities.
Using
open-source
Video
and
Image
Analytics
Marine
Environment
(VIAME),
we
trained
neural
network
model
identify
frames
containing
krill.
Our
image
classifier
has
overall
accuracy
73%,
with
positive
predictive
value
83%
prediction
then
developed
method
estimate
volume
water
imaged,
thus
(N·m
-3
krill,
2-dimensional
images.
The
is
based
on
maximum
range
camera
where
remain
visibly
resolvable
assumes
that
mean
length
known,
distribution
orientation
angles
uniform.
From
1,932
identified
as
manually
subset
124
across
record
contained
unresolvable
necessary
imaged
sensor.
Krill
swarm
penguins
ranged
2
307
krill·m
was
48
(sd
=
61
).
Mean
biomass
25
g·m
.
frame-level
estimation
provide
new
efficiently
process
video-logger
data
2D
imagery,
providing
key
information
aggregations
may
affect
predator
performance.
should
be
directly
applicable
other
feeding
prey.
Three-dimensional
(3D)
imaging
has
attracted
more
and
interests
because
of
its
widespread
applications,
especially
in
information
life
science.
These
techniques
can
be
broadly
divided
into
two
types:
ray-based
wavefront-based
3D
imaging.
Issues
such
as
quality
system
complexity
these
limit
the
applications
significantly,
therefore
many
investigations
have
focused
on
from
depth
measurements.
This
paper
presents
an
overview
measurements,
provides
a
summary
connection
between
techniques.
Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE,
Journal Year:
2016,
Volume and Issue:
9713, P. 97131T - 97131T
Published: March 9, 2016
A
novel
spectral
imaging
technique
is
introduced
based
on
a
highly
dispersive
lens
system.
The
chromatic
aberration
of
the
system
utilized
to
spread
content
object
over
focal
distance.
Two
three-dimensional
surface
reconstruction
algorithms,
depth
from
focus
and
defocus,
are
applied
images
captured
by
Using
these
imager
able
relate
either
location
focused
image
or
amount
defocus
at
detector
object.
with
~5
nm
resolution
designed
this
technique.
spatial
resolutions
independent
can
be
improved
simultaneously.
Simulation
experimental
results
presented.
In
this
paper,
we
propose
an
end-to-end
self-supervised
Deep
Neural
Network
(DNN)
for
Defocus
Map
Estimation
(DME).
Currently,
such
defocus
maps
are
estimated
by
DNNs
with
fully
supervised
learning.
For
training,
networks
need
large
datasets
annotated
amount
or
scene
depth,
which
challenging
to
obtain.
The
training
arrive
overcome
the
limitation
represented
Ground
Truth
(GT)
data.
line,
a
learning
neural
network
DME
from
single
defocused
image.
Our
method
is
based
on
recently
proposed
DNN
called
2HDED:NET
that
enriched
simulation
module
makes
possible
DME.
addition
map,
our
reconstructs
All-in-Focus
(AIF)
image
through
We
test
synthetic
and
realistic
benchmarks
demonstrate
it
effective
solution
deblurring
when
available.
2021 2nd International Conference on Smart Electronics and Communication (ICOSEC),
Journal Year:
2021,
Volume and Issue:
unknown, P. 918 - 921
Published: Oct. 7, 2021
Games
have
evolved
from
the
beginning
of
human
civilization
to
modern
times.
One
oldest
and
most
popular
are
role-playing
games,
which
allow
player
enjoy
playing
a
role
in
game
play
variety
characters.
In
this
paper,
similarity
redundancy
3D
object
information
used,
CS
technology
is
combined
improve
computational
efficiency,
while
alleviating
high
energy
consumption
sampling
transmission
process
diffraction
field.
whole
experiment,
paper
uses
visualization
tools
analyze
summarize
data
experiment.
