TENCON 2021 - 2021 IEEE Region 10 Conference (TENCON),
Год журнала:
2022,
Номер
unknown, С. 1 - 6
Опубликована: Ноя. 1, 2022
Imaging
under
photon-scarce
situations
introduces
challenges
to
many
applications
as
the
captured
images
are
with
low
signal-to-noise
ratio.
Here,
we
target
on
denoising
photon-limited
imaging.
We
develop
a
multi-level
pyramid
network
(MPDNet),
which
employs
idea
of
Laplacian
learn
small-scale
noise
map
and
larger-scale
high-frequency
details
at
different
levels.
Feature
extractions
conducted
multi-scale
input
encode
richer
contextual
information.
The
major
component
MPDNet
is
multi-skip
attention
residual
block,
integrates
feature
fusion
mechanism
for
better
representation.
Experimental
results
have
demonstrated
that
our
can
achieve
superior
performance
photon
counts.
2022 26th International Conference on Pattern Recognition (ICPR),
Год журнала:
2022,
Номер
unknown
Опубликована: Авг. 21, 2022
Light
field
(LF)
images
can
record
the
scene
from
multiple
directions
and
have
many
applications,
such
as
refocusing
depth
estimation.
However,
these
applications
be
heavily
influenced
by
poor
light
condition
noise.
This
work
aims
to
recover
high-quality
LF
their
lowlight
detection.
First,
a
decomposition
network
is
employed
decompose
each
image
into
its
reflectance
illumination
with
Retinex
theory.
Then,
two
enhancement
networks
are
designed
denoise
enhance
illumination,
respectively.
They
adopt
alternate
spatial-angular
feature
extractions
process
all
views
synchronously
high
efficiency.
A
parallel
dual
attention
mechanism
integrated
both
spatial
angular
extractions,
encode
more
important
information.
Moreover,
discriminator
introduced
during
training
generate
realistic
making
judgment
according
characteristics.
Experimental
results
demonstrated
superior
performance
of
our
method,
which
restore
content,
luminance,
color
geometric
structures
effectively.
arXiv (Cornell University),
Год журнала:
2021,
Номер
unknown
Опубликована: Янв. 1, 2021
Imaging
under
photon-scarce
situations
introduces
challenges
to
many
applications
as
the
captured
images
are
with
low
signal-to-noise
ratio
and
poor
luminance.
In
this
paper,
we
investigate
raw
image
restoration
low-photon-count
conditions
by
simulating
imaging
of
quanta
sensor
(QIS).
We
develop
a
lightweight
framework,
which
consists
multi-level
pyramid
denoising
network
(MPDNet)
luminance
adjustment
(LA)
module
achieve
separate
enhancement.
The
main
component
our
framework
is
multi-skip
attention
residual
block
(MARB),
integrates
multi-scale
feature
fusion
mechanism
for
better
representation.
Our
MPDNet
adopts
idea
Laplacian
learn
small-scale
noise
map
larger-scale
high-frequency
details
at
different
levels,
extractions
conducted
on
input
encode
richer
contextual
information.
LA
enhances
denoised
estimating
its
illumination,
can
avoid
color
distortion.
Extensive
experimental
results
have
demonstrated
that
restorer
superior
performance
degraded
various
photon
levels
suppressing
recovering
effectively.
arXiv (Cornell University),
Год журнала:
2023,
Номер
unknown
Опубликована: Янв. 1, 2023
Depth
estimation
from
light
field
(LF)
images
is
a
fundamental
step
for
numerous
applications.
Recently,
learning-based
methods
have
achieved
higher
accuracy
and
efficiency
than
the
traditional
methods.
However,
it
costly
to
obtain
sufficient
depth
labels
supervised
training.
In
this
paper,
we
propose
an
unsupervised
framework
estimate
LF
images.
First,
design
disparity
network
(DispNet)
with
coarse-to-fine
structure
predict
maps
different
view
combinations.
It
explicitly
performs
multi-view
feature
matching
learn
correspondences
effectively.
As
occlusions
may
cause
violation
of
photo-consistency,
introduce
occlusion
prediction
(OccNet)
maps,
which
are
used
as
element-wise
weights
photometric
loss
solve
issue
assist
learning.
With
estimated
by
multiple
input
combinations,
then
fusion
strategy
based
on
errors
effective
handling
final
map
accuracy.
Experimental
results
demonstrate
that
our
method
achieves
superior
performance
both
dense
sparse
images,
also
shows
better
robustness
generalization
real-world
compared
other
TENCON 2021 - 2021 IEEE Region 10 Conference (TENCON),
Год журнала:
2022,
Номер
unknown, С. 1 - 6
Опубликована: Ноя. 1, 2022
Imaging
under
photon-scarce
situations
introduces
challenges
to
many
applications
as
the
captured
images
are
with
low
signal-to-noise
ratio.
Here,
we
target
on
denoising
photon-limited
imaging.
We
develop
a
multi-level
pyramid
network
(MPDNet),
which
employs
idea
of
Laplacian
learn
small-scale
noise
map
and
larger-scale
high-frequency
details
at
different
levels.
Feature
extractions
conducted
multi-scale
input
encode
richer
contextual
information.
The
major
component
MPDNet
is
multi-skip
attention
residual
block,
integrates
feature
fusion
mechanism
for
better
representation.
Experimental
results
have
demonstrated
that
our
can
achieve
superior
performance
photon
counts.
