What Does It Means Automatic Understanding of the Images?

Abstract

Images become very important source of information. Very often when we have both text and image part in a document, the image contains most important part of information. Example of such situation can be medical documentation of some patient. Medical images (e.g. MRI or PET presentations) are more informative for diagnosis and for therapy control than descriptions given in text form. Therefore when we have to find proper document in multimedial database as well as we must made some analysis (eg. statistical) based on stored material -we must go into the image merit content. Searching and analysis methods based on figure captions and another descriptions given in text form are definitely not enough for such purposes. Moreover the typical or traditional forms of computer vision methods, like image processing, analysis and also pattern recognition are not sufficient, because all of them are based on image forms (shapes, textures etc.) when we must in fact go to the image merit content. In the paper the problem of automatic understanding of the images is formulated, discussed, and solved.