FINGER DETECTION IN VIDEO SEQUENCES USING A NEW SPARSE REPRESENTATION
Vasile Gui, Daniel Popa, Pekka Nisula, Veijo Korhonen · 2011
In this paper we propose a new method for finger detection in video sequences. The method is robust and has a low computational cost. From a background subtracted image, we generate a sparse image representation, based on line strip features. We use a robust and adaptive clustering method, related to the mean shift (MS) to detect and track fingers, as well as to extract finger position parameters from validated clusters. In contrast with the traditional mean shift tracker, based on color histogram, we use shape information to detect and track hand fingers. Our experiments prove that the tracker is accurate and robust to extreme noise tests and partial occlusion. We applied the proposed method in a human computer interface designed to control very large public displays.