Restoring a silhouette of the hand in the problem of recognizing gestures by adaptive morphological filtering of a binary image

Algorithms are presented for the adaptive processing of binary images of silhouettes of the human hand obtained by means of color-brightness filters. These algorithms are based on the use of a combination of elementary morphological operations that take into account the direction of the fingers. Algorithms are presented for removing noise on binary images that are adapted to the result of the operation of a color filter, and a method is presented for filling internal contours of a silhouette of the hand in order to remove grouped marking errors. The experimental results show that the proposed image-processing method increases the probability of successful detection, tracking of the hand, and recognition of gestures. (C) 2014 Optical Society of America.