Stereo matching on uncalibrated images based on epipolar geometry
基于极线几何约束的非标定图像的立体匹配

Stereo matching is a key issue in the area of computer vision, and the core issue of the 3D reconstruction of computer as well. It plays a very important role in feature matching and is independent of scene structure. Combining feature matching with epipolar geometry, a robust matching algorithm was carried out on uncalibrated images. The least epipolar distance theory was applied to the fundamental matrix estimate and guided match. The results of the experiments on real images show that this algorithm has better practicality.