Photogrammetric approach fo1r precise correction of camera misalignment for 3D image generation

Abstract

This paper proposes a new algorithm for correcting camera misalignments and removing Y-parallax between stereo images. Traditionally, this task has been tackled by estimating the fundamental matrix directly from tie points and applying epipolar resampling based on the fundamental matrix. We propose a photogrammetric approach for the task: Instead of direct estimation of the fundamental matrix, we estimate relative orientation between stereo images, calculating the fundamental matrix from the relative orientation and applying epipolar resampling. We compared the performance of the proposed algorithm with the traditional one. We found that the proposed one could generate epipolar resampled images with less distortions and could be more effective for 3D image generation.