Local Invariant Object Localization based on Gabor Feature Space

R. Heintz, G. Schäfer, and G. Bretthauer (Germany)


Object recognition, Gabor filter, robust local features, rotation and scale invariant, object localization, feature space


Invariant object localization is one of the challenging tasks in computer vision research. In this paper we present a robust rotation and scale invariant object localization method. A local Gabor filter space is treated as core of this method. Two dimensional rotation and scaling operations were transformed in shift operations along the Gabor filter space dimensions. This property enables efficient scale and rotation estimation without segmentation. The method was tested with two standardized image databases and with images out of an industrial environment. Besides a good quality in localizing rotated and scaled objects, the method has a strong robustness against variations of lightning conditions and 3D viewpoint changes.

Important Links:

Go Back