Abstract
Unlike textual databases, image databases are fairly new. This is largely because image processing is a highly computationally intensive activity. Storage overhead for an image database is also considerable. Despite these problems significant progress has already been made because of advances in VLSI technology. This paper will describe how we successfully harnessed the parallelism offered by a transputer network, which is a product of VLSI technology, in providing the necessary processing power for creating an image database. The database consists of a collection of historical museum photographs. It is implemented on an IBM compatible PC equipped with the necessary image processing and mass storage facilities and linked to a network of four transputers. The user interface is provided through a serial mouse and a menu of keywords. Users will be able to retrieve individual images in three different modes: (i) the browse mode will allow the user to display a set of up to 12 photographs in miniature “stub” format at one go: a click on a stub will select it for display in full size, (2) the search mode involves the selection of a keyword from the menu of subject or location index in order to retrieve a specified image, (3) users can also retrieve a photograph by typing-in the unique museum reference number. The paper will also discuss the results of our image processing experiments—enhancement, compression and decompression—very briefly.