A Study on the Planar Rectification of Self-Calibrated Stereo Images

Abstract

In this paper, we study a planar rectification technique that takes perspective projection matrices obtained by self-calibration. The self-calibration takes a nonlinear estimation called the Levenberg-Marquardt method on constraints imposed by the fundamental matrix to estimate the projection matrices. The rectification technique determines new perspective projection matrices by solving four ^(3x4)linear homogeneous systems obtained from rectifying plane constraints. This paper shows experimental results obtained by applying the technique to controlled stereo data and real stereo images. We make the comparison based on the average vertical coordinate difference of the rectified stereo image pair.

In this paper, we study a planar rectification technique that takes perspective projection matrices obtained by self-calibration. The self-calibration takes a nonlinear estimation called the Levenberg-Marquardt method on constraints imposed by the fundamental matrix to estimate the projection matrices. The rectification technique determines new perspective projection matrices by solving four ^(3x4)linear homogeneous systems obtained from rectifying plane constraints. This paper shows experimental results obtained by applying the technique to controlled stereo data and real stereo images. We make the comparison based on the average vertical coordinate difference of the rectified stereo image pair.