Binarising SIFT-Descriptors to Reduce the Curse of Dimensionality in Histogram-Based Object Recognition (bibtex)
by Martin Stommel, Otthein Herzog
Abstract:
It is shown that distance computations between SIFT-descriptors using the Euclidean distance su er from the curse of dimensionality. The search for exact matches is less a ected 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 SIFT-descriptors is a much better choice.
Reference:
Martin Stommel, Otthein Herzog, "Binarising SIFT-Descriptors to Reduce the Curse of Dimensionality in Histogram-Based Object Recognition", In SIP, Springer, no. 61, Jeju Island, Korea, pp. 320-328, 2009.
Bibtex Entry:
@INPROCEEDINGS{Stommel2009,
  author = {Stommel, Martin and Herzog, Otthein},
  title = {Binarising SIFT-Descriptors to Reduce the Curse of Dimensionality
	in Histogram-Based Object Recognition},
  booktitle = SIP,
  year = {2009},
  editor = {Slezak, Dominik and Pal, Sankar K. and Kang, Byeong-Ho and Gu, Junzhong
	and Kurada, Hideo and Kim, Tai-hoon},
  number = {61},
  series = {Communications in Computer and Information Science},
  pages = {320--328},
  address = {Jeju Island, Korea},
  month = {December10--12},
  publisher = {Springer},
  abstract = {It is shown that distance computations between SIFT-descriptors using
	the Euclidean distance suer from the curse of dimensionality. The
	search for exact matches is less aected 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 SIFT-descriptors is a much better choice.},
  isbn = {ISSN 1865-0929},
  owner = {pmania},
  timestamp = {2012.11.06}
}
Powered by bibtexbrowser