Image annotation is a time-consuming and costly task. Previously, we published MorphoClusteras a novel image annotation tool to address problems of conventional, classifier-basedimage annotation approaches: their limited efficiency, training set bias and lack of novelty detection.MorphoCluster uses clustering and similarity search to enable efficient, computer-assisted imageannotation. In this work, we provide a deeper analysis of this approach. We simulate the actions ofa MorphoCluster user to avoid extensive manual annotation runs. This simulation is used to testsupervised, unsupervised and transfer representation learning approaches. Furthermore, shrunkenk-means and partially labeled k-means, two new clustering algorithms that are tailored specificallyfor the MorphoCluster approach, are compared to the previously used HDBSCAN*. We find thatlabeled training data improve the image representations, that unsupervised learning beats transferlearning and that all three clustering algorithms are viable options, depending on whether completeness,efficiency or runtime is the priority. The simulation results support our earlier findingthat MorphoCluster is very efficient and precise. Within the simulation, more than five objects persimulated click are being annotated with 95% precision.