MapReduce Vocabulary Tree: An Approach for Large Scale Image Indexing and Search in the Cloud

摘要

In this article we present a novel approach, based on the MapReduce paradigm, for the construction of Vocabulary Trees designed for addressing large scale image indexing and search problems. The proposed approach encompasses two stages: (i) key-points detection and feature extraction from images, and (ii) vocabulary tree construction. An experimental evaluation generated speedup and scalability figures that demonstrate the behavior of the proposed approach. Accuracy figures and image retrieval examples on some datasets are also presented.