J. Richarz, T. Pl{\"o}tz and G. A. Fink
Proc. IASTED Int. Conf. on Visualization, Imaging and Image Processing
(VIIP 2007), pages 259-264, 2007.
Palma de Mallorca, Spain
In this paper, we describe our approach on hand detection in cluttered images using scale invariant features. We claim that, while modelling hands as a whole is bound to fail because of their strongly articulated nature, treating them as a collection of weakly connected characteristic regions seems promising. Different approaches to finding and robustly modelling such regions - or local object descriptors - invariantly to scale and orientation of the object in question have been proposed. As an example, we demonstrate our approach using the well-known scale-invariant feature transform (SIFT), combined with a region-based postprocessing to eliminate false positives. We present detailed results on a large set of images from a realistic interaction scenario with a smart room.