Databases Reference
In-Depth Information
7)
The most correlated image (semantically closest image) to the given context
as the user's impression (Kansei information) is extracted in the selected
subspace.
The semantic associative search system selects appropriate images for user's requests
with Kansei information by using metadata items and basic functions. This system
consists of the following subsystems:
(1)
Image Selection Subsystem: This subsystem supports the facilities for
selecting appropriate images by using the mathematical model of meaning.
Three methods are provided for representing the metadata items for images.
This subsystem maps the metadata items in the semantic space created in
the step 2). When the context is given as the Kansei information, this
subsystem selects the most correlated image to the context.
(2)
Metadatabase Management Subsystem: This subsystem supports the
facilities for keeping metadata consistent in the orthogonal semantic space.
(3)
Metadata Acquisition Subsystem: This subsystem supports the facilities for
acquiring metadata from the database storing the source images by receiving
user's requests with Kansei information.
1.3
Basic functions and metadata for images
The metadatabase system is used to extract image dta items corresponding to
context words which represent the user's impression and the image's contents. For
example, the context words “powerful” and “strong” are given, the image with
the impression corresponding to these context words is extracted. Each metadata
item of images is mapped in the orthonormal semantic space. This space is referred
to as “orthogonal metadata space” or “metadata space.” The MMM is used to
create the orthogonal metadata space. By this orthogonalization, we can define
appropriate metric for computing relationships between metadata items for images
and context representation. The MMM gives the machinery for extracting the
associated information to the context.
Three types of metadata are used.
(1)
Metadata for space creation:
These metadata items are used for the creation
of the orthogonal metadata space, which is used as a space for semantic
image retrieval.
(2)
Metadata for images:
These metadata items are the candidates for semantic
image retrieval.
(3)
Metadata for context representation:
These metadata items are used as
context words for representing the user's imagination and the image's
contents.
The basic functions and metadata structures are summarized as follows:
