Depth Extraction System Using Stereo Pairs

Abstract

. Stereo vision refers to the ability to infer information on the 3-D structure of a scene from two or more images taken from different viewpoints. Stereo pairs stand as an imperative source for depth extraction. The same are being used for giving vision to automated vehicles and creating Digital Elevation Models from satellite stereo pairs. Processing a pair involves certain steps and techniques exist for each one. But there is not a completely defined approach which encompasses these steps directing to depth extraction from an uncaliberated stereo pair. This paper describes a system which automatically recovers the depth information form two narrow baseline frames. Our contributions include the development of the system, automation of the complete process, comparison of the rectification approaches and the stereo matching techniques for disparity maps. We have proposed a system which takes the uncaliberated stereo pair as the input and gives estimated relative depth. After testing a number of stereo pairs and comparing them with the ground truth, the experimental results demonstrate that our proposed approach leads to a system encompassing the state-of-art algorithms which extracts the relative depth information from a stereo pair.