The formulation of polyurethanes is a complex poorly understood problem and it has developed more as an art rather than a science. Although polyurethane formulations can be developed from first principles, this approach requires both a detailed knowledge of the underlying principles that govern the formulation process and also time, since a number of measurements of process conditions are usually required. The case-based reasoning (CBR) methodology can support polyurethane formulation tasks by providing a framework for collecting, structuring, and representing historical formulating knowledge. To date, most CBR retrieval algorithms employ a modified version of the nearest neighbour rule that uses a distance function as similarity measure, which in turn depends upon the attribute type. The application of moment-based retrieval used in image recognition for CBR retrieval is studied in this paper. Comparison with the classical retrieval algorithms that use standard distance measures showed that low-order geometric, central, and Legendre moments retrieve the same cases as the Euclidean distance does, whereas high-order geometric, central, and Legendre moments retrieved different cases. It is suggested that there is not a single distinguished approach to similarity in CBR, rather CBR systems should allow the integration of different approaches to similarity and the selection of different concepts.