real time stereo matching using back propagation neural network and hierarchical method

in this paper a fast and reliable method is introduced for real time stereo matching problem such as intelligent vehicle applications. In this method vertical edges of stereo images are completely extracted using two levels of threshold. Then for each edge number of edge points distance of two end points edge angle edge sign and destance between mean points of edge points and middle point of two end points are extracted as features for each two edges of the left and right images that we want to compare them together difference of these features are computed and will be inputs of a back propagation neural network. neural network was trained using a supervised learning method.

Keywords:

feature based stereo mathing , back propagation neural network , fast cross correlation , two levels of threshold , hierarchical stereo matching