Image indexing - A First Author Co-citation Map

By Olle Persson
Dept sociology
Umeå univ
901 87 Umeå
Sweden

E-mail: Olle.Persson@soc.umu.se

Date: 2002-01-01

The map below is an attempt to define the knowledge base of image indexing. Using Web of Science (WoS) I started with a review paper by Edie Rasmussen: Indexing images, Annual Review of Information Science and Technology (ARIST), 32(1997)169-196. Then another 100 papers were selected that share references with the Rasmussen paper in descending order of similarity. Thus, these papers have similar reference lists and can be assumed to discuss the same topic. Among the cited papers the most cited first authors were selected. Then a matrix of co-citations were built which was taken as input to a MDS-algoritm that produced the coordinates for the map. The circle size indicates the number of citations. The closer the authors are the more frequently they are co-cited. Please note that some names may have variant spelling, for example "Shatford" and "Layne" are the same person.

Here is how Edie Rasmussen reads the map:
"In the ARIST chapter I identify two basic approaches to indexing images, concept-based and content-based. In concept-based indexing, images and objects therein are manually identified and described in terms of what they are and what they represent, and in content-based indexing certain features of the images, such as colour, are automatically identified and extracted. As Cawkell had earlier pointed out, there is a lack of interrelationships between groups with these two approaches, with the former being centered in the imaging and library and information science community, and the latter in the computer science community. This division is illustrated by the co-citation map. The lower group in the map (Svenonius, Lunin, Markey, Besser, Rorvig, etc.) are all part of the library and information science community; the appearance of Panofsky in this group reflects his role as art historian in providing the theoretical grounding for the concept-based approach with which this group is identified. In the co-citation map, the LIS group is isolated from the much larger computer-science group (e.g. Jain, Srihari, Chang, Faloutsos) who have focussed on automatic or content-based approaches. Interestingly, Salton is central to this group; though he did not to my knowledge work on multimedia retrieval, he is associated with automatic approaches to text retrieval. The visual confirmation of Cawkell's original observation is quite interesting! (A.E. Cawkell (1992). Selected aspects of image processing and management: review and future prospects. Journal of Information Science 18(3): 179-192.)"

Try click on (some) names