Selection of local features for visual search

In a compact descriptor for visual search only a limited number of local features may be included. The estimated probability for correct match between keypoints provides a good criterion for selection of a subset.

► Selection of a subset of SIFT features enables small queries for visual search.
► The selected SIFT features are those estimated as most likely to be correctly matched.
► The DoG parameters used are peak, scale, orientation and location.
► The method ensures high recognition rates even at small query sizes.