Image
Retrieval: Theory and Research
by Corinne Jörgensen
Scarecrow Press, Inc.. Lanham, Maryland,
and Oxford, 2003
360 pp. Trade, $47.00
ISBN: 0-8108-4734-5.
Reviewed by Pia Tikka
Researcher in University of Art and Design
Hämeentie 135 C, 00560 Helsinki,
Finland
Pia.Tikka@uiah.fi
Image Retrieval: Theory and Research
is a book, which serves, like its author
Corinne Jörgensen emphasizes, as
a foundation for individuals in research
communities, who need an introductory
overview on the issues of image access,
indexing, and retrieval. It becomes clear
along reading this excellent book that
it is intended for experienced readers:
not only taken the quite fast "first-aid"
coverage of the exhaustive fields that
relate to image retrieval and systems
design, such as the domains of cognitive
psychology, the psychology of art, computer
science, content-based image retrieval
(CBR), and library and information science
(LIS), but also due to the compactness
of the information presented.
It is made clear that the cross-disciplinary
user-centered approach is a must
in future research. The large body of
literature concerning data models, pattern
and feature recognition, database architectures,
and machine learning and image understanding
through applications of artificial intelligence,
Jörgensen notes, have concentrated
on what is computationally possible, but
not on the analysis of the needs of real-world
users. The support outside of the scope
of the book could be found in Andy Clarks
[1] discussion on the inseparable reciprocal
evolutionary relationship with the human
mind and techno- environment, i.e. "specially
constructed environments replete with
artifacts, external symbols, and all the
variegated scaffoldings of science, art
and culture". Jörgensen suggests
unifying research technology, methods
and goals that facilitate "a common,
sensible, and easily understood framework"
(p. 199) for the future interdisciplinary
research community.
Jörgensens preliminary research
interest lies on identifying and categorizing
image attributes in user-queries and image-indexing
[2]. The research draws from the human
tendency for "similarity" and
"critical difference" judgment,
which provides plausible tools for image
feature categorizing. The perceptually
stimulated attributes, such as the human
figure, objects, color, and location,
are typically recognized in images. Jörgensens
research concludes that the contextual
situatedness of each individual, be it
a user or an indexer, fundamentally enables
or constrains the image classification
and management processes. Emotional moods,
personal motivations, likes and dislikes,
life-experience etc. affect the descriptions
and feature attributes given to the images
when indexing, and, in turn, to the information
requested when searching images. Thus,
an ecological view [3] surfaces towards
the end of the book.
Jörgensens wish list for the
future research involves following: a
testbed for comparative evaluation,
which would incorporate large number of
images from diverse domains (science,
arts, consumers etc.); a visual indexing
language for bridging the gap between,
for example, the automated lower-level
feature-based image-content analyses and
the higher-level semantic content of images;
an integrated indexing record accommodating
multiple types of data from both human
and machine process; widening communities
of research and practice juxtaposing
the researchers experiences and
interests from the domains of the CBR
and the LIS. Final but not least in the
researchers wish list is a request
for a new starting point: Understanding
the needs of an end-user and her desired
methods of interaction should guide creation
of "real-world" applications,
namely usable, flexible interfaces.
The critical notion of the content
of the book draws from the apparent two-facet
problem of visual and textual information.
Though the book emphasizes the research
orientation towards the integration of
visual and textual knowledge, it deliberately
ignores this in the content of the book
itself. In my view, both for less experienced
readers and for researches not familiar
with the computational image-analyzing
systems, the access to the image indexing
and retrieval topic might have become
more feasible with some visualization.
Use of images could have added to the
discussion the actual dimension that the
authors approach emphasizes, the
visual knowledge and embodied understanding
as an unexplored resource of image retrieval
systems design.
In concluding, I recommend the book as
an excellent introductory text to the
field of image processing. Jörgensens
overview on image retrieval, i.e. human-generated
and machine-processed practices and theory,
makes an important contribution for the
designers of a globally accessible image
management. It seems, though, that the
success or failure of the future research
fundamentally relies on the application
of what the Western science has become
to know as embodied knowledge.
References:
[1] Clark, A., "Reasons, Robots and
the extended Mind". In: Mind &
Language, Vol. 16 No. 2 April 2001:
pp. 121-145. (p. 142).
[2] Jörgensen, C., "Image Attributes:
An Investigation", Ph.D. thesis,
Syracuse University, 1995.
[3] With ecological approach is referred
to the idea that the organism evolves
in continuous interaction with its environment.
Here the reviewer assumes that the author
relates with the view of Gibson, J.J.
1979. The ecological approach to visual
perception. Houghton Mifflin, Boston.