SIFTpack: a compact representation for efficient SIFT matching
Alexandra Gilinsky
Technion
sashkagil@gmail.com
Technion
lihi@ee.technion.ac.il
Why SIFTpack?
Abstract
Computing distances between large sets of SIFT descriptors is a basic step in numerous algorithms in computer vision. When the number of descriptors is large, as is often the case, computing these distances can be extremely time consuming. In this paper we propose the SIFTpack: a compact way of storing SIFT descriptors, which enables significantly faster calculations between sets of SIFTs than the current solutions. SIFTpack can be used to represent SIFTs densely extracted from a single image or sparsely from multiple different images. We show that the SIFTpack representation saves both storage space and run time, for both finding nearest neighbors and for computing all distances between all descriptors. The usefulness of SIFTpack is demonstrated as an alternative implementation for K-means dictionaries of visual words and for image retrieval.
Key idea
Results
The key contribution of this work is a significant reduction in run-time when computing distances between sets of SIFTs. We suggest solutions for exact and approximate nearest-neighbor matching and all-distances calculation. The proposed solutions are also shown to be useful for bag-of-words models in terms of run-time and memory space.
Paper
Download
The code in this website is for demo purposes only. Individuals or academic institutes are free to use this version as long as they acknowledge its use. Commercial licensing is managed by the Technion Industry Liaison Office. Please contact Hovav Gazit for details.
Acknowledgements
This research was supported in part by the Ollendorf foundation and by the Israel Ministry of Science. We'd also like to thank Prof. Michael Elad for useful conversations and good ideas.