In computer vision applications where the calibration object is not avaible , it is useful to use an uncalibrated stereoscopic head. Even in this case, to calculate the three-dimensional structure of the viwed scene, the stereo matching is considered as the key step in stereo vision analysis. This paper presents a contribution to resolve this problem when an uncalibrated stereo rig is involved in a visual task. We propose an algorithm for self-matching of stereoscopic images of indoor scenes. Based on projective geometry, the principal idea of the method is to estimate the epipole position assuming a set of matched 2D surfaces. A voting approach is used to select the correct matching which produce the same solution. In practice, as the stereo images are noisy, we propose a mathematical analysis of the uncerainty measure. We assume that the vertices are noisy, and we propagate the effect of this noise in the different stages of the proposed algorithm. The new version of the algorithm allows to calculate the region where the epipole point appertains, called the "epipolar region". The stereo matching algorithm has been tested on both synthetic and real images, and the number of lines matched demostrates the robustness of the geometric method.
JavaScript jest wyłączony w Twojej przeglądarce internetowej. Włącz go, a następnie odśwież stronę, aby móc w pełni z niej korzystać.