Databases Reference
In-Depth Information
4.2
Method-2
The image
P
is represented in t impression words
o
1
,
o
2
,…,
o
t
,
where each impression
word is defined as a t dimensional vector:
which is characterized by n specific features.
Namely, we define the image
P
as the collection of t impression words which
represent the image.
Moreover, we define the operator union of impression words
o
1
,
o
2
,…,
o
t
, to
represent the metadata for the image
P
as a vector as follows:
where sign
(a)
represents the sign (plus or minus) of
“a”
and
l
k
, k
=1,…,
n,
represents
the index which gives the maximum, that is:
4.3
Method-3
In this method, the metadata for images is automatically and indirectly extracted
from image data items themselves. Color is known as the dominant factor which
affects the impression of images
10,
11,
1)
. We use color to derive impressions of images.
The basic idea of this method is to describe both images and impression words
in the notion of
color
and compute correlations between images and words. Color
used in this method is represented in the Munsell color system as it is more familiar
to the human perception. Additionally, the color names defined by ISCC(Inter-Society
Color Council) and NIST(National Institute of Standards and Technology) are used
to describe both images and impression words in the notion of color. By taking the
correlations between images and impression words, we can obtain the suitable words
which describe the impressions of images. The metadata for images is computed
from the obtained impression words by the previously defined union operator
.
5
Examples of Creating Metadata
5.1
Method-1
To create the metadata as vectors for images by using the 850 basic words of
“General Basic English Dictionary,” the designer of metadata looks at an image
