It is a challenge for today’s library practice to make the entities in digital images retrievable according to the users’ demands. In our paper we wish to introduce a new search method which is able to seek certain elements and their environment in the images by applying within them a search in a natural language. One key element of this method is to extend the Dublin Core metadata system by a new qualifier of the description element which helps to enlist the natural language denominations of the elements in the images and the elements’ positions within it. Subsequently, we are able to find certain elements, then calculate and grade the information value of their environment with the help of a modified version of Shannon’s entropy formula. The application of this method enables us to perform further value calculations for certain details within a particular image which makes, further, more efficient searches possible.