Comparing two- and three-view Computer Vision
This work addresses incremental improvements in 3D reconstruction techniques for computer vision applications, with limited broader impact.
The paper compared two-view and three-view computer vision methods for 3D point reconstruction, finding that the two-view method reconstructs significantly more points, while the three-view method yields smaller dispersion, with both methods performing similarly under varying camera settings.
To reconstruct the points in three dimensional space, we need at least two images. In this paper we compared two different methods: the first uses only two images, the second one uses three. During the research we measured how camera resolution, camera angles and camera distances influence the number of reconstructed points and the dispersion of them. The paper presents that using the two-view method, we can reconstruct significantly more points than using the other one, but the dispersion of points is smaller if we use the three-view method. Taking into consideration the different camera settings, we can say that both the two- and three-view method behaves the same, and the best parameters are also the same for both methods.