Method and apparatus for determining a sharpness metric of an image
In order to objectively and efficiently measure, visualize and evaluate the quality of images in terms of sharpness, a method for determining a sharpness metric (S) of an image is described. The method includes performing edge detection that results in gradients for the pixels of the image, determining a value representing a contrast metric C of the image, calculating at least three different thresholds (t 1 ,t 2 ,t 3 ) from the image content, and classifying the pixels according to their gradient into at least four groups defined by the thresholds. The sharpness metric (S) is calculated from relationships between the amounts of pixels in the different groups and a multiplicative factor (m) between the at least three thresholds.
1. A method for determining a sharpness metric (S) of an image (im), comprising
performing edge detection on the image, wherein gradient values are assigned to pixels;
determining a value representing a contrast metric C of the image;
calculating at least a first threshold t 1 , a second threshold t 2 and a third threshold t 3 from the determined contrast metric C, wherein the first, second and third thresholds are related by a constant factor m according to t 1 =C, t 2 =m*t 1 and t 3 =m*t 2 , wherein m is greater than one;
determining an amount of pixels having a gradient value of at least the first threshold t 1 , the pixels being first level edge pixels (B 1 ),
an amount of pixels having a gradient value of at least the second threshold t 2 , the pixels being second level edge pixels (B 2 ), and
an amount of pixels having a gradient value of at least the third threshold t 3 , the pixels being third level edge pixels (B 3 ); and
determining the sharpness metric S from the factor m, a relationship S 2 between the amounts of second level and first level edge pixels and a relationship S 3 between the amounts of third level and first level edge pixels.
2. The method according to claim 1 , wherein the sharpness metric S is calculated according to
S
=
S
2
+
(
m
-
1
)
S
3
m
=
S
2
-
S
3
m
+
S
3
,
with
S
2
=
∑
B
2
/
∑
B
1
and
S
3
=
∑
B
3
/
∑
B
1
wherein B 1 are the first level edge pixels, B 2 are the second level edge pixels and B 3 are the third level edge pixels.
3. The method according to claim 1 , wherein the contrast metric C is a weighted Root-Mean-Square (RMS) value (C WRMS ) over pixels of the image, wherein pixels that according to the edge detection belong to an edge have higher weight than other pixels.
4. The method according to claim 3 , wherein the weighted RMS value (C WRMS ) is calculated as a square root of a normalized weighted average of square deviations from a mean value L mean according to
C
WRMS
=
∑
W
(
L
-
L
mean
)
2
∑
W
where L denotes luminance values of pixels, W denotes weights obtained in said edge detection and being assigned to the pixels, and L mean is a sum of weighted luminance values divided by a sum of luminance values.
5. The method according to claim 3 , wherein the weighted RMS value (C WRMS ) is scaled with a constant factor greater than one to obtain a scaled weighted RMS value (C WRMS,s ), wherein the clipped scaled weighted RMS value (C WRMS,s ) is said contrast metric C of the image.
6. The method according to claim 5 , wherein the scaled weighted RMS value (C WRMS,s ) is clipped to a maximum value of C max =L max −L min with L max and L min being maximum and minimum luminance values respectively of pixels within the image, wherein the clipped scaled weighted RMS value is said contrast metric C of the image.
7. The method according to claim 3 , wherein the edge detection provides a gradient g for a pixel, and the weight w of the pixel is calculated according to w=g 2 for obtaining said weighted Root-Mean-Square (RMS) value (C WRMS ).
8. The method according to claim 1 , wherein the constant factor m is in a range of [1.9, . . . , 2.1].
9. The method according to claim 1 , further comprising
storing the sharpness metric C and assigning it as metadata to the image.
10. The method according to claim 1 , further comprising
calculating a fourth threshold t 4 being related to the third threshold by said constant factor m according to t 4 =m*t 3 ;
determining an amount of pixels having a gradient value of at least the fourth threshold t 4 , the pixels being fourth level edge pixels (B 4 ); and
determining an oversharpness metric S 4 being a relationship between the amount of fourth level edge pixels and the amount of first level edge pixels.
11. The method according to claim 1 , further comprising
performing image processing based on the sharpness metric S, wherein pixels of the image are modified.
12. The method according to claim 1 , further comprising
controlling a camera based on the sharpness metric S (to correct image sharpness), wherein at least one of an exposure time, an aperture and an image sensing parameter of the camera is controlled.
13. An apparatus for determining a sharpness metric (S) of an image (im), comprising
an weight generator unit adapted for generating weights for pixels and assigning the weights to pixels;
a contrast metric determining unit adapted for determining a value representing a contrast metric C of the image;
a calculating unit adapted for calculating at least a first threshold t 1 , a second threshold t 2 and a third threshold t 3 from the determined contrast metric C, wherein the first, second and third thresholds are related by a constant factor m according to t 1 =C, t 2 =m*t 1 and t 3 =m*t 2 , wherein m is greater than one;
a counter unit adapted for determining an amount of first level edge pixels (B 1 ) being pixels having a gradient value of at least the first threshold t 1 ,
determining an amount of second level edge pixels (B 2 ) being pixels having a gradient value of at least the second threshold t 2 , and
determining an amount of third level edge pixels (B 3 ) being pixels having a gradient value of at least the third threshold t 3 ; and
a calculation unit adapted for determining the sharpness metric S from the factor m, a relationship S 2 between the amounts of second level and first level edge pixels and a relationship S 3 between the amounts of third level and first level edge pixels.
14. The apparatus according to claim 13 , wherein the sharpness metric S is calculated according to
S
=
S
2
+
(
m
-
1
)
S
3
m
=
S
2
-
S
3
m
+
S
3
,
with
S
2
=
∑
B
2
/
∑
B
1
and
S
3
=
∑
B
3
/
∑
B
1
wherein B 1 are the first level edge pixels, B 2 are the second level edge pixels and B 3 are the third level edge pixels.
15. The apparatus according to claim 13 , wherein the contrast metric C is a weighted Root-Mean-Square (RMS) value (C WRMS ) over pixels of the image, wherein pixels that according to the edge detection belong to an edge have higher weight than other pixels.