Local object description using Gabor features
Joni-Kristian Kamarainen
(U. of Lappeenranta, Finland)
Recently, local object descriptors have been one of the most active
research areas in computer vision. Their importance partly yields from
the fact that structural variance in visual appearance of real objects
and object categories can be efficiently modeled using a connected
structure, spatial constellation, of local descriptors. Representing
local visual information is also an intrinsic property of features
constructed from Gabor filter responses; That was actually the
original motivation in the first computer vision related publication
of Gabor features in 1978 - claiming Gabor filter as "general image
processing operator". This talk will cover the fundamentals of feature
extraction using Gabor filters, fast implementations, important
properties of Gabor features, and finally, demonstrate their use as
local object descriptors.