An algorithm is given for processing pictorial query specifications that consist of a query image and a similarity level that must hold between the query image and database images. The similarity level specifies the contextual similarity (how well the content of one image matches that of another) as well as the spatial similarity (the relative locations of the matching symbols in the two images). The algorithm differs from previous approaches in its ability to handle multiple instances of each object in both the query and database images by searching for isomorphic subgraphs. The running time of the algorithm is O(m 2m) in the worst case where all symbols in both the query and database image are from the same class, but falls far below this bound in the presence of spatial constraints.