Cultural memory of images has traditionally linked images with texts, terms and verbal indexes. Confronted with the transition of images into digital storage gradually non-verbal methods of classification gain importance. The question arises which new kind of knowledge will exists exclusively in the form of images, which part of traditional knowdledge can be transformed into images and which part might just vanish.

The research group "Image archives", consisting of the programmer and artist Stefan Heidenreich (Berlin/Cologne), the art historian and philosopher Peter Geimer (Berlin) and the media archaeologist and historian Wolfgang Ernst (Berlin/Cologne) in cooperation with the Academy of Media Arts (Cologne) aims at rethinking the notion of images from the vantage point of the process of archiving. The archive here is seen as a medium of storage und form of organization of all that can be accessed as knowledge. The function of archives of images such as museums or data banks exceeds by far mere storage and conservation of images. Instead of just collecting passively and subsequently archives actively define what is meant to the archivable at all. In so far they determine as well what is allowed to be forgotten.

In terms of technology an archive is a coupling of storage media, format of contents and address structure. In this case the images is to be conceived as data format. Methododically this implies leaving behind the contemplation and description of single images in favour of an investigation of sets of images.

In his 1766 essay "Laocoon" G. E. Lessing discussed the media conflict between the logic of language and the logic of images in terms of cultural semiotics. The digitalization of images today provides a technical basis of inquiry into this conflict, so that this investigation can be grounded in the terms of the medium computer. It would not make sense to retell a teleological story of images processing which finally reached its aim in digitalization; on the contrary this history of images is to be revised from the present point of view of digitalization. How can, for example, archives be related to algorithms of image processing, of pattern recognition and computer graphics?

In sharp contrast to hermeneutics the present investigation of image archives do not take images as carriers of experiences and meanings. The relation between vision and image cannot be taken as the guideline of investigation, since image processing by computers can no more be re-enacted with the anthropological semantics of the human eye. The methodological starting point of research is rather a theory of technical media based on Michel Foucault´s discourse analysis and Claude Shannon's mathematical theory of communication, as well as practices and notions of data-structure oriented programming.

The following topics of research are going to be focussed:

• Artes memoriae: visual techniques of memorization from the rhetorics of the antique to the renaissance.
• Museums, collections, images of picture galleries, catalogues: Programming of material image banks
• Visual knowledge in eighteenth century: Encyclopedias and their visualizations
• Photography as switching medium from perception to technology: Pre-visual image archives
• Movies as archives: Hollywood and the rules of image sequences
• Digital image archives: Text- and imageoriented structures in image data bases

Related to these fields of research a selection of unpublished documents on methods of image archivization relevant in cultural history shall be edited. The work of the research group shall not be limited to the investigation of given images archives, but deals with programming data bases. Priority has the development of a visually adressable image archive. By combining Multiresolutional Image Representation with simple Octree structures a variable archive module shall be provided. This allows to test the application of different algorithms creating different visual sequences and neighbourhoods. Most operators of image processing and pattern recognition such as filters and invariant transformations can be integrated in the structure of the data base in order to make cluster of images accessable. A scientific practice of dealing with images will thus not only be reflected in terms of media thery, but as well be designed as software.

The devolpment of such a program is not dependent on a previously determined application. Based on the assumption that media have an effect on their contents rather than vice versa the aim is not to find solutions to already given problems. The reflection of the technical basis and of the structure of image data will lead to the development of an algorithm which only then will show the way to possible applications.

The next stept will be the development of an interactive and visual agent capable of "intelligent" retrieval of images and visual sketches in large data. In order to involve culturally accustomed ways of relating images in the retrieval and in the visualization of the user interface a detailed investigation of modes of perception throughout cultural history is of central importance.