Fast Exact Nearest Neighbour Matching in High Dimensions Using -D Sort
Figure 2
Results for SIFT descriptor data experiment using euclidean distance. (a) Search structure construction time plots against database size. (b) Exact 1NN query time, normalised according to linear search time, plots against database size. A threshold of 2.0 was used. (c) Exact 1NN query time, normalised according to linear search time, plots against threshold for 100 k descriptor database. (d) Exact query time, normalised according to linear search time, plots against the number of neighbours to find. A 100 k descriptor database and threshold of 2.0 was used. (e) Accuracy of approximate 1NN query plots against query time, normalised according to linear search time. A 100 k descriptor database and threshold of 2.0 were used.
