Real-time disparity computation and tracking for hand-pair gesture recognition

This thesis presents an algorithm for the real-time computation of disparity using video stereo images and for real-time tracking of objects. This algorithm is used as part of a hand-pair gesture recognition application. The first part of the thesis covers the modification to the OpenCV stereo algorithm so that it is made to run at more than 20 frames per second for the hand-pair gesture recognition application. The second part of the thesis implements a FIR polynomial filter as a prediction filter that is used to fill in the samples in the hand-pair gesture recognition application, when no segmentation is done. The results obtained provide the performance of the developed algorithm running in real-time.