A stereo depth recovery method using belief propagation

Abstract

Stereo vision refers to the ability to infer information on the 3D structure of scene from two or more images taken from different viewpoints. This paper describes procedure for depth map creating using rectified stereo images and segmentation algorithm belief propagation (BP). Very necessary steps to creating depth map are camera calibration and image rectification of the image pairs. Calibration of the stereoscopic cameras consists from a two parameters of a stereo system: intrinsic parameters, which characterize the transformation mapping an image point from camera to pixel coordinates in each camera and extrinsic parameters, which describe the relative position and orientation of the two cameras. The depth recovery is important problem of image analysis in the study of compute vision and is optimized by using the belief propagation techniques.