Method for calculating the depth map from a stereo pair

Abstract

Active and passive methods of restoring information about the depth of a real scene are known. Active methods use laser illumination of the working space, giving the output fast and accurate information about the depth . However, these methods have limitations with respect to the measurement range and cost of hardware components. Passive methods based on computer vision are usually implemented with simpler and less expensive distance sensors. Such methods are able to generatecameras . One of the fundamental problems of stereo vision is to establish an exact match between the left and right image of the stereo pair. Correspondence refers to the distance (disparity) between pixels of the same object on the left and right image. This paper describes a passive method for calculating the depth map in real time on a GPU.