The application of a qualitative shape and qualitative colour description and similarity calculus to an Image Query By Example problem is presented in this paper. Specifically, the qualitative shape and colour similarity theories are applied to the problem of matching icon images. The suitability of this approach as a foundation for a practical query by example engine for icons is shown by an experimental evaluation. The use of a more cognitive approach opens the possibility of creating systems that can be tailored to particular user requirements involving complex rules that cannot be handled by current systems.