System of the multicriterional identification of the trade marks images

Abstract

In the present paper the experimental system for the automated search of images in a database with elements of comparative estimations is announced. The initial data for the recognition system are stylized graphic images of trade marks. Identified images concerning own a particular standard can have various scales, rotated, insignificant linear and brightness distortions. The presented approach is concerned with representing a complex composition (scene) by elementary fragments (objects). To recognize the given scene all the objects are singled out on the image, are quantitatively estimated separately and as a complexe scene concerning some reference point.