Clustering Similar Images Using MapReduce Style Feature Extraction with C# and R

The createPairwiseMatches() function shown in Figure 7 above, extracts features in parallel mapping images to vertical and horizontal luminosity histograms. Furthermore, the histograms for each image are saved in a hash table for quick reference since each image's features will be repeatedly matched to other images. Once the match features are extracted, the match is immediately placed in a thread safe blocking collection for further downstream reduction processing. While the mapping functions shown in Figure 7 are executing in a background thread, parallel reduce functions simultaneously execute processing each completed match produced to calculate the similarity between the match images.