Imagistic domains require from the problem-solver the ability of applying primarily visual recognition of objects, and only secondarily search and analytical methods in order to interpret these objects. Most of the time, the visual recognition process is developed at the sensorial level and has no related conceptualisations. This work explores the use of ontologies and introduces the concept of visual chunk, as a modelling primitive applied to elicit and represent visual objects in complex domains. A problem-solving method is defined to describe the reasoning process of object recognition and interpretation. Based on this method, a full knowledge system application was developed and is being used to support the interpretation of sedimentary rocks through petrographic analysis.