Survey of SIFT Compression Schemes

Transmission and storage of local feature descriptors are of critical importance for mobile visual search applications. We perform a comprehensive survey of Scale Invariant Feature Transform (SIFT) compression schemes proposed in the literature and evaluate them in a common framework. Further, we compare the different schemes to the recently proposed low bit-rate Compressed Histogram of Gradients (CHoG) descriptor. We show that CHoG outperforms all SIFT compression schemes. We implement CHoG in a large-scale mobile image retrieval system and show that transmitting CHoG feature data are an order of magnitude smaller than transmitting SIFT descriptors or JPEG images.