Abstract
The rapid increase in the volume of archived of image and video materials has given rise to an urgent need for new methods of image retrieval. However, to be useful such methods must be both specific enough to respond to image structure which is like to be salient, and flexible enough to exhibit substantial invariance to irrelevant transformations. It may therefore be expected that global aspects as retrieval criteria. We propose a retrieval criterion which is a compound of global features with respective weights which may be tailored to suit the target material, and we give examples of its performance.