Increasing interest in configuration similarity is currently developing in the context of Digital Libraries, Spatial Databases and Geographical Information Systems. The corresponding queries retrieve all database configurations that match an input description (e.g. 'find all configurations where an object x0 is about 5 km north-east of another x1, which, in turn, is inside object x2'). This paper introduces a framework for configuration similarity that takes into account all major types of spatial constraints (topology, direction, and distance). We define appropriate fuzzy similarity measures for each type of constraint to provide flexibility and allow the system to capture real-life needs. Then we apply preprocessing techniques to explicate constraints in the query, and present algorithms that effectively solve the problem. Extensive experimental results demonstrate the applicability of our approach to images and queries of considerable size.