Method for segmenting an image
A method for isolating an element of an image made up of pixels comprising the steps of classifying the pixels into different groups based on the color value of the pixel, blurring the image, locating a pixel in the blurred image that has a predetermined color value corresponding to the element to be isolated, and growing a mask from the located pixel.
1. A method for isolating an element of an image made up of pixels comprising the steps of classifying the pixels into different groups based on the color value of the pixel, blurring the image, locating a pixel in the blurred image that has a predetermined color value corresponding to the element to be isolated, and growing a mask from the located pixel.
2. The method of claim 1 wherein the step of classifying the pixels into groups is performed using a K-means algorithm.
3. The method of claim 2 wherein the step of growing the mask is performed by adding to the located pixel the pixels in the group in which the located pixel is classified that are spatially connected to the located pixel using a connection algorithm.
4. The method of claim 3 wherein the color of a pixel in the mask is adjusted.
5. The method of claim 4 wherein prior to the step of classifying the image, the image is in RGB color space and is converted to the CIELAB color space.
6. The method of claim 5 wherein the step of converting the image to the CIELAB color space is modeled as a linear system
t=G T c
where the pixel value is given by the 3-element vector c, t is the 3-element CIEXYZ value, and G is a 3×3 matrix that is determined by solving
G
=
arg
(
min
H
(
1
N
∑
i
=
1
N
F
(
H
T
c
i
)
-
u
i
)
)
where the CIELAB values are given by the vector sequence {u i } i=1 N , the RGB values are given by {c i } i=1 N , and the mapping from CIEXYZ to CIELAB is given by the function F.
7. The method of claim 6 wherein in the step of classifying the pixels includes increasing the dynamic range of the image.
8. The method of claim 7 wherein the image is blurred by using a convolution process including a blur kernel and the blur kernel is smaller than the feature to be isolated.
9. The method of claim 8 wherein the convolution process is given by the equation
B
(
I
(
x
,
y
)
)
=
∫
-
∞
∞
∫
-
∞
∞
H
(
a
,
b
)
I
(
x
-
a
,
y
,
b
)
ⅆ
a
ⅆ
b
where H(a,b) is the blur kernel.
10. The method of claim 9 wherein the function H is given by the equation
H ( a,b )=1/(2 N +1) 2 ∀−N≦a≦N & − N≦b≦N
where N is the radius of the blur kernel.
11. The method of claim 9 wherein the step of locating the pixel uses the algorithm
[
x
0
,
y
0
]
=
arg
(
min
[
a
,
b
]
(
B
(
I
(
a
,
b
)
)
-
v
2
)
)
where the color of the expected feature value is given by the three element vector v.
12. The method of claim 2 is wherein the number of groups is 3 to 5.
13. The method of claim 2 wherein the number of groups is sufficient to differentiate the element to be isolated from other elements of the image.
14. The method of claim 2 wherein the step of classifying the pixels includes refining the color value of the different groups based on the group average and reclassifying the pixels in the image based on the refined color values.
15. A computer-readable medium containing instructions for controlling a processor to isolate an element in an image made up of pixels by a method comprising the steps of classifying the pixels into different groups based on the color value of the pixel, blurring the image, locating a pixel in the blurred image that has a predetermined color value corresponding to the element to be isolated, and growing a mask from the located pixel.
16. The medium of claim 15 wherein the step of classifying the pixels into groups is performed using a K-means algorithm.
17. The medium of claim 16 wherein the step of growing the mask is performed by adding to the located pixel the pixels in the group in which the located pixel is classified that are spatially connected to the located pixel using a connection algorithm.
18. The medium of claim 17 wherein the color of a pixel in the mask is adjusted.
19. The medium of claim 18 wherein prior to the step of classifying the image, the image is in RGB color space and is converted to the CIELAB color space.
20. The medium of claim 16 wherein there are 3 to 5 groups.