Image Processing Reference
In-Depth Information
(4.5a) Sample image categories
(4.5b) Sample image categories
FIGURE 4.5: Image universe and categories
, some sam-
ples from flowers category. the set C
contains 100 flower images
(4.6a) The subspace
C
(4.6b) original query image,
A
(4.6c) lower approximation in
terms of the universe X
FIGURE 4.6: Obtaining lower approximation of a query image in terms of the universe
three color components as features
categories. This will reveal important information about the degree of similarity between
the query image and the categories. We define an equivalence relation on the set of images
X. Three color components in addition to an image index are considered as features:
ϕ(p
ij
) = (ϕ
(p
ij
), ϕ
(p
ij
), ϕ
(p
ij
), ϕ
(p
ij
))
1
2
3
4
ϕ
1
(p
ij
) = R(p
ij
),
ϕ
(p
ij
) = G(p
ij
),
2
ϕ
(p
ij
) = B(p
ij
),
3
ϕ
4
(p
ij
) = i, i = 1, 2, . . . , j = 1, 2, . . . , M
i
,
where p
ij
is the j
th
pixel of i
th
image(I
i
) in the universe; M
i
is the number of pixels in
the image I
i
. R extracts the red component of the pixel, G extracts green and B, blue
component. Equivalence classes are formed on each image, I∈X. So each image I is
partitioned into equivalence classes based on features defined by ϕ, I
/∼,ϕ
.
Search WWH ::
Custom Search