Stereo Image Rectification Using Focal Length Adjustment

Abstract

Stereo Image rectification is more popular due to its robustness and vast amount of search space reduction in stereo correspondence problem. It has wide range of applications in 3d medical imaging, robot navigation, virtual reality, and entertainment etc. Rectified Images includes more distortion which will affect 3d imaging quality. In this paper, we choose time stamped stereo frames from stereo video streams from heterogeneous cameras. These stereo image pairs are verified for suitability for calibration using error pair elimination method. Heterogeneous cameras are calibrated from the selected stereo image pairs and followed by rectification of stereo image pairs by compensating the focal lengths of the cameras. The resultant rectified image has lesser number of hollow pixels as compared to conventional algorithm. Lesser number of hollow pixels results in accurate 3D co-ordinate estimation compared with conventional algorithm.