4.1 Local Maxima:
We use Local maxima which were detected by observing
each image point in SIFT descriptor. A point is decided as a
local minimum or maximum when its value is smaller or
larger than all its surrounding neighboring points by a certain
amount as used by [13] after SIFT extraction stage. Because
the feature obtained by SIFT was large number we used
locale maxima to reduce the time needed to image database
process faster, even the result might not changed.
Bernd apply feature reduction to the top level classifie
choosing relevant image features according to a measure
derived fromstatistical learning theory [17].