Replica detection is an important prerequisite for the discovery of copyright infringement and detection of illicit content. For this purpose, content-based image protection can be an efficient alternative to watermarking. Rather than imperceptibly embedding a signal, content-based systems rely on image similarity. Certain content-based systems use binary classifiers to detect replicas, each classifier being fine-tuned to a particular original. However, since a suspect image has to be tested against every original, such a comparison becomes computationally prohibitive as the number of original images grows. In this paper, we propose an indexing method to efficiently prune the number of comparisons with the originals in the database. For this purpose, a multidimensional indexing structure, namely R-trees, is incorporated to rapidly select the most likely originals. Experimental results showed that up to 97% of the database can be discarded before applying the binary classifiers.