Binarising SIFT-Descriptors to Reduce the Curse of Dimensionality in Histogram-Based Object Recognition (bibtex)

by Stommel, Martin

Abstract:

It is shown that distance computations between SIFT-descriptors using the Euclidean distance suffer from the curse of dimensionality. The search for exact matches is less affected than the generalisation of image patterns, e.g. by clustering methods. Experimental results indicate that for the case of generalisation, the Hamming distance on binarised SIFTdescriptors is a much better choice. It is shown that the binary feature representation is visually plausible, numerically stable and information preserving. In an histogram-based object recognition system, the binary representation allows for the quick matching, compact storage and fast training of a code-book of features. A time-consuming clustering of the input data is redundant.

Reference:

Stommel, Martin, "Binarising SIFT-Descriptors to Reduce the Curse of Dimensionality in Histogram-Based Object Recognition", In IJSIP, vol. 3, no. 1, pp. 25–36, 2010. Publisher: Science & Engineering Research Support Center SERSC

Bibtex Entry:

@ARTICLE{Stommel2010,
  author = {Stommel, Martin},
  title = {Binarising SIFT-Descriptors to Reduce the Curse of Dimensionality
	in Histogram-Based Object Recognition},
  journal = {IJSIP},
  year = {2010},
  volume = {3},
  pages = {25--36},
  number = {1},
  month = {March},
  note = {Publisher: Science {\&} Engineering Research Support Center SERSC},
  abstract = {It is shown that distance computations between SIFT-descriptors using
	the Euclidean distance suffer from the curse of dimensionality. The
	search for exact matches is less affected than the generalisation
	of image patterns, e.g. by clustering methods. Experimental results
	indicate that for the case of generalisation, the Hamming distance
	on binarised SIFTdescriptors is a much better choice. It is shown
	that the binary feature representation is visually plausible, numerically
	stable and information preserving. In an histogram-based object recognition
	system, the binary representation allows for the quick matching,
	compact storage and fast training of a code-book of features. A time-consuming
	clustering of the input data is redundant.},
  issn = {ISSN: 2005-4254},
  keywords = {Object recognition,curse of dimensionality,SIFT,binarisation,clustering},
  owner = {pmania},
  timestamp = {2012.11.06}
}