Method and device for removing haze in single image
View Patent ↗The present invention provides a method for removing a haze in a single image. In the present invention, a transmission is estimated by using a dark channel prior obtained from a hazy input image. The estimated transmission includes a block artifact. In an exemplary embodiment of the present invention, in order to preserve an edge and remove the block artifact, a refined transmission value is obtained by performing WLS filtering by using an estimated transmission value and a morphologically-processed input image, the image is restored based on the refined transmission value, and then multi-scale tone manipulation image processing is performed.
1. A device for removing a haze in a single image, the device comprising:
a processor; and
a non-transitory computer-readable recording medium storing instructions which, when executed by the processor, cause the processor to perform operations comprising:
calculating a first transmission by estimating a transmission value based on a dark channel prior generated from a hazy input image,
calculating a second transmission for minimizing a block artifact generated in an image restoration process based on the first transmission by applying weighted least square (WLS) filtering based on the first transmission and morphologically-processed input image information,
generating a restored image by removing a haze component in the hazy input image based on the second transmission, and
performing multi-scale tone manipulation image processing on the restored image,
wherein the second transmission is calculated by an equation,
∑
p
(
(
r
p
-
r
~
p
)
2
+
λ
(
w
x
,
p
(
h
)
(
∂
r
∂
x
)
p
2
+
w
y
,
p
(
h
)
(
∂
r
∂
y
)
p
2
)
)
,
where p denotes a position of each pixel in the hazy input image, r p denotes a refined transmission, {tilde over (r)} p denotes the estimated transmission value, λ denotes a smoothing control constant, h denotes a morphologically-processed hazy image, w x,p (h) and w y,p (h) respectively denote horizontal-axis and vertical-axis smoothing weight values of the morphologically-processed hazy image, and
(
∂
r
∂
x
)
p
and
(
∂
r
∂
y
)
p
respectively denote horizontal-axis and vertical-axis first-order differential operators of r.
2. The device of claim 1 , wherein the dark channel prior is generated by an equation
J
dark
(
p
)
=
min
q
∈
Ω
(
p
)
(
min
c
∈
{
r
,
g
,
b
}
J
c
(
q
)
)
,
J
dark
→
0
where J dark denotes a dark channel prior; J c denotes an input image of each color channel; and Ω denotes a block around a P pixel.
3. The device of claim 1 , wherein the second transmission is a value obtained by preserving edge information and removing a block artifact in the first transmission based on the edge information of the morphologically-processed input image.
4. The device of claim 1 , wherein the multi-scale tone manipulation image processing is performed by dividing the restored image into a base layer, a medium layer, and a fine layer.
5. A method for removing a haze in a single image, the method comprising:
calculating, by a transmission estimating unit, a first transmission by estimating a transmission value based on a dark channel prior generated from a hazy input image;
calculating, by a transmission refining unit, a second transmission for minimizing a block artifact generated in an image restoration process based on the first transmission by applying weighted least square (WLS) filtering based on the first transmission and morphologically-processed input image information;
generating, by an image restoring unit, a restored image by removing a haze component in the hazy input image based on the second transmission; and
performing, by an image processing unit, multi-scale tone manipulation image processing on the restored image,
wherein the second transmission is calculated by an equation,
∑
p
(
(
r
p
-
r
~
p
)
2
+
λ
(
w
x
,
p
(
h
)
(
∂
r
∂
x
)
p
2
+
w
y
,
p
(
h
)
(
∂
r
∂
y
)
p
2
)
)
,
where p denotes a position of each pixel in the hazy input image, r p denotes a refined transmission, {tilde over (r)} p denotes the estimated transmission value, λ denotes a smoothing control constant, h denotes a morphologically-processed hazy image, w x,p (h) and w y,p (h) respectively denote horizontal-axis and vertical-axis smoothing weight values of the morphologically-processed hazy image, and
(
∂
r
∂
x
)
p
and
(
∂
r
∂
y
)
p
respectively denote horizontal-axis and vertical-axis first-order differential operators of r p .
6. The device of claim 1 , wherein the first transmission is calculated by an equation,
r
~
(
p
)
=
1
-
μ
min
q
∈
Ω
(
p
)
(
min
c
∈
{
r
,
g
,
b
}
I
c
(
q
)
A
c
)
wherein p denotes a pixel position of each pixel in the hazy input image, {tilde over (r)}(p) denotes the estimated transmission value for the pixel position p, μ denotes a constant, Ω(p) denotes a block around the pixel position p, c denotes a color channel, r denotes a red color channel, g denotes a green color channel, b denotes a blue color channel, I denotes the hazy input image, and A denotes a global atmospheric light source.
7. The device of claim 1 , wherein the smoothing weight values w x,p (h), w y,p (h) of the morphologically-processed hazy image are used to preserve a region corresponding to an edge region of the morphologically-processed hazy image and smooth another region thereof in the estimated transmission, and are respectively obtained with equations,
w
x
,
p
(
h
)
=
(
∂
h
∂
x
p
α
+
ɛ
)
-
1
,
w
y
,
p
(
h
)
=
(
∂
h
∂
y
p
α
+
ɛ
)
-
1
,
where α denotes an edge control constant of the morphologically-processed hazy image h that is implemented to further smooth the another region other than the region corresponding to the edge region of the morphologically-processed hazy image as α increases; and ε denotes a constant value for preventing a denominator from being divided by 0.
8. The device of claim 1 , wherein r p is further smoothed as λ increases.