Automatic sharpening of images
View Patent ↗Various embodiments describe image sharpening techniques that automatically estimate a sharpening amount for an unsharp mask filter for image enhancement based upon a statistical correlation between detail coefficients at a first resolution and at a second resolution. In various embodiments, statistical deviation of absolute values of detail coefficients for both the image at full resolution and the image downsampled by a pre-determined factor, e.g. a factor of two (half resolution), are retrieved. In various embodiments, the statistical deviation is retrieved for a histogram of the image at the first resolution and a histogram of the image at the second resolution. The linear model between the statistical deviation of detail coefficients at the first resolution and the second resolution is used to calculate the sharpening amount and the unsharp mask filter is applied to the image to produce a sharpened image.
1. A computer-implemented method comprising:
receiving an image having an initial resolution;
automatically calculating a sharpening amount based on the image by utilizing at least one calculated statistical deviation of a histogram of the image at the initial resolution and at least one calculated statistical deviation of a histogram of the image at a second resolution;
applying an unsharp mask filter to the image, the unsharp mask filter employing the sharpening amount and being represented by an expression
X′=AF ( X−G ( X ))+ X
where X′ is an adjusted value of a pixel in the image, X is a value of the pixel in the image before sharpening, A is the sharpening amount, and G(X) is a result of a convolution of X with a Gaussian kernel G with a standard deviation of one pixel, and where
F
(
x
)
=
x
if
x
≥
τ
and
F
(
x
)
=
x
2
τ
otherwise
where τ is a positive threshold; and
outputting a sharpened image relative to the image received.
2. The computer-implemented method of claim 1 , the second resolution of the image being one-half of the initial resolution of the image.
3. The computer-implemented method of claim 1 , wherein the at least one calculated statistical deviation calculated comprises one or more of a standard deviation, an absolute deviation, a median absolute deviation, a third-quartile absolute deviation, a ninth-decile absolute deviation, or a ninety-ninth-centile absolute deviation.
4. The computer-implemented method of claim 1 , automatically calculating a sharpening amount based on the image by utilizing at least one calculated statistical deviation of a histogram of the image at the initial resolution and at least one calculated statistical deviation of a histogram of the image at a second resolution comprising:
generating a histogram of absolute values of detail coefficients of the image at the initial resolution;
calculating at least one statistical deviation of the histogram of the image at the initial resolution;
generating a histogram of absolute values of detail coefficients of the image at the second resolution;
calculating at least one statistical deviation ˜f the histogram of the image at the second resolution; and
calculating the sharpening amount according to an expression
A
=
clamp
(
0
,
A
max
,
min
(
μ
D
1
,
D
max
)
D
0
-
1
)
where
clamp
(
v
min
,
v
max
,
value
)
=
max
(
v
min
,
min
(
v
max
,
value
)
)
and where A is the sharpening amount, D 1 is the statistical deviation of the histogram of the image at the second resolution, D 0 is the statistical deviation for histogram of the image at the initial resolution, D max is a predetermined maximum value for the statistical deviation, μ is equal to 0.94, and A max is equal to 2.0.
5. The computer-implemented method of claim 4 , wherein the at least one statistical deviation comprises one or more of a standard deviation, an absolute deviation, a median absolute deviation, a third-quartile absolute deviation, a ninth-decile absolute deviation, or a ninety-ninth centile absolute deviation.
6. One or more computer-readable storage media comprising instructions that are executable to cause a device to perform a process comprising:
automatically calculating a sharpening amount based on an image by utilizing at least one statistical deviation of a histogram of the image at a first resolution and at least one statistical deviation of a histogram of the image at a second resolution by
calculating two or more of a standard deviation, an absolute deviation, a median absolute deviation, a third-quartile absolute deviation, a ninth-decile absolute deviation, or a ninety-ninth-centile absolute deviation of the histogram of the image at the initial resolution;
comparing the two or more calculations to determine a conservative statistical deviation of the histogram of the image at the initial resolution;
calculating a conservative statistical deviation of the histogram of the image at the second resolution, the conservative statistical deviation of the histogram of the image at the second resolution corresponding to the conservative statistical deviation of the histogram of the image at the initial resolution; and
calculating the sharpening amount according to an expression
A
=
clamp
(
0
,
A
max
,
min
(
μ
D
1
,
D
max
)
D
0
-
1
)
where
clamp
(
v
min
,
v
max
,
value
)
=
max
(
v
min
,
min
(
v
max
,
value
)
)
and where A is the sharpening amount, D 1 is the conservative statistical deviation of the histogram of the image at the second resolution, D 0 is the conservative statistical deviation for histogram of the image at the initial resolution, D max is a predetermined maximum value for the statistical deviation, μ is equal to 0.94, and A max is equal to 2.0;
applying an unsharp mask filter to the image, the unsharp mask filter employing the sharpening amount; and
outputting a sharpened image relative to the image.
7. A device comprising:
one or more processors;
one or more computer-readable storage media; and
one or more modules embodied on the one or more computer-readable storage media and executable under the influence of the one or more processors, the one or more modules comprising an image sharpening module configured to perform operations comprising:
receiving an image having an initial resolution;
automatically calculating a sharpening amount based on the image by utilizing at least one calculated statistical deviation of a histogram of the image at the initial resolution and at least one calculated statistical deviation of a histogram of the image at a second resolution;
applying an unsharp mask filter to the image, the unsharp mask filter employing the sharpening amount, wherein the unsharp mask filter is represented by an expression
X′=AF ( X−G ( X ))+ X
where X′ is an adjusted value of a pixel in the image, X is a value of the pixel in the image before sharpening, A is the sharpening amount calculated, and G(X) is a result of a convolution of X with a Gaussian kernel G with a standard deviation of one pixel, and where
F
(
x
)
=
x
if
x
≥
τ
and
F
(
x
)
=
x
2
τ
otherwise
where ρ is a positive threshold; and
outputting a sharpened image relative to the image received.
8. The device of claim 7 , wherein the second resolution of the image is one half of the initial resolution of the image.
9. The device of claim 7 , wherein the at least one calculated statistical deviation calculated comprises one or more of a standard deviation, and absolute deviation, a median absolute deviation, a third-quartile absolute deviation, a ninth-decile absolute deviation, or a ninety-ninth-centile absolute deviation.
10. The device of claim 7 , wherein said automatically calculating comprises:
generating a histogram of absolute values of detail coefficients of the image at the initial resolution;
calculating at least one statistical deviation of the histogram of the image at the initial resolution;
generating a histogram of absolute values of detail coefficients of the image at the second resolution;
calculating at least one statistical deviation of the histogram of the image at the second resolution; and
calculating the sharpening amount according to an expression
A
=
clamp
(
0
,
A
max
,
min
(
μ
D
1
,
D
max
)
D
0
-
1
)
where
clamp
(
v
min
,
v
max
,
value
)
=
max
(
v
min
,
min
(
v
max
,
value
)
)
and where A is the sharpening amount, D 1 is the statistical deviation of the histogram of the image at the second resolution, D 0 is the statistical deviation for histogram of the image at the initial resolution, D max is a predetermined maximum value for the statistical deviation, μ is equal to 0.94, and A max is equal to 2.0.
11. The device of claim 10 , wherein the at least one statistical deviation calculated comprises one or more of a standard deviation, an absolute deviation, a median absolute deviation, a third-quartile absolute deviation, a ninth-decile absolute deviation, or a ninety-ninth-centile absolute deviation.
12. A computer-implemented method comprising:
receiving an image having an initial resolution;
automatically calculating a sharpening amount based on the image by utilizing at least one calculated statistical deviation of a histogram of the image at the initial resolution and at least one calculated statistical deviation of a histogram of the image at a second resolution by:
generating a histogram of absolute values of detail coefficients of the image at the initial resolution;
calculating at least one statistical deviation of the histogram of the image at the initial resolution;
generating a histogram of absolute values of detail coefficients of the image at the second resolution;
calculating at least one statistical deviation ˜f the histogram of the image at the second resolution; and
calculating the sharpening amount according to an expression
A
=
clamp
(
0
,
A
max
,
min
(
μ
D
1
,
D
max
)
D
0
-
1
)
where
clamp
(
v
min
,
v
max
,
value
)
=
max
(
v
min
,
min
(
v
max
,
value
)
)
and where A is the sharpening amount, D 1 is the statistical deviation of the histogram of the image at the second resolution, D 0 is the statistical deviation for histogram of the image at the initial resolution, D max is a predetermined maximum value for the statistical deviation, μ is equal to 0.94, and A max is equal to 2.0;
applying an unsharp mask filter to the image, the unsharp mask filter employing the sharpening amount; and
outputting a sharpened image relative to the image received.
13. The computer-implemented method of claim 12 , wherein the at least one statistical deviation comprises one or more of a standard deviation, an absolute deviation, a median absolute deviation, a third-quartile absolute deviation, a ninth-decile absolute deviation, or a ninety-ninth centile absolute deviation.
14. The computer-implemented method of claim 12 , the unsharp mask filter being represented by an expression
X′=AF ( X−G ( X ))+ X
where X′ is an adjusted value of a pixel in the image, X is a value of the pixel in the image before sharpening, A is the sharpening amount, and G(X) is a result of a convolution of X with a Gaussian kernel G with a standard deviation of one pixel, and where
F
(
x
)
=
x
if
x
≥
τ
and
F
(
x
)
=
x
2
τ
otherwise
where τ is a positive threshold.
15. The computer-implemented method of claim 12 , the unsharp mask filter being represented by an expression
X′=AF ( X−G ( X ))+ X
where X′ is an adjusted value of a pixel in the image, X is a value of the pixel in the image before sharpening, A is the sharpening amount, and G(X) is a result of a convolution of X with a Gaussian kernel G with a standard deviation of one pixel, and where
F ( x )= x if | x |≧τ and F ( x )=0 otherwise
where τ is a positive threshold.
16. A device comprising:
one or more processors;
one or more computer-readable storage media; and
one or more modules embodied on the one or more computer-readable storage media and executable under the influence of the one or more processors, the one or more modules comprising an image sharpening module configured to perform operations comprising:
receiving an image having an initial resolution;
automatically calculating a sharpening amount based on the image by utilizing at least one calculated statistical deviation of a histogram of the image at the initial resolution and at least one calculated statistical deviation of a histogram of the image at a second resolution by:
generating a histogram of absolute values of detail coefficients of the image at the initial resolution;
calculating at least one statistical deviation of the histogram of the image at the initial resolution;
generating a histogram of absolute values of detail coefficients of the image at the second resolution;
calculating at least one statistical deviation of the histogram of the image at the second resolution; and
calculating the sharpening amount according to an expression
A
=
clamp
(
0
,
A
max
,
min
(
μ
D
1
,
D
max
)
D
0
-
1
)
where
clamp
(
v
min
,
v
max
,
value
)
=
max
(
v
min
,
min
(
v
max
,
value
)
)
and where A is the sharpening amount, D 1 is the statistical deviation of the histogram of the image at the second resolution, D 0 is the statistical deviation for histogram of the image at the initial resolution, D max is a predetermined maximum value for the statistical deviation, μ is equal to 0.94, and A max is equal to 2.0;
applying an unsharp mask filter to the image, the unsharp mask filter employing the sharpening amount; and
outputting a sharpened image relative to the image received.
17. The device of claim 16 , wherein the at least one statistical deviation calculated comprises one or more of a standard deviation, an absolute deviation, a median absolute deviation, a third-quartile absolute deviation, a ninth-decile absolute deviation, or a ninety-ninth-centile absolute deviation.
18. The device of claim 16 , wherein the unsharp mask filter is represented by an expression
X′=AF ( X−G ( X ))+ X
where X′ is an adjusted value of a pixel in the image, X is a value of the pixel in the image before sharpening, A is the sharpening amount calculated, and G(X) is a result of a convolution of X with a Gaussian kernel G with a standard deviation of one pixel, and where
F
(
x
)
=
x
if
x
≥
τ
and
F
(
x
)
=
x
2
τ
otherwise
where τ is a positive threshold.