Currently, the main challenge of CBIR is bridging the so-called semantic gap. A question emerges whether visual similarity is reliable for grading semantic similarity and how complex the problem of semantic interpretation of an image really is? Using the UML class diagram, a model for semantic interpretation of general images is presented. A general model for still images of outdoor and indoor scenes is defined, independent of domain the image belongs to and the objects it contains. Analysis and visual presentation of key model elements aid in observing the problem in whole and simplify the process of finding new solutions.