New Methods for 3D Scene Creation Using 2D Images

Two of multiple view object reconstruction approaches are inspected in this work. The goal is to create and test several functions for processing of two input images of the scene to uncover geometry of the scene. A new algorithm for Maximally Stable Extremal Regions correspondence detection is presented – True Tentative Correspondences employing the Sideness constraint. The input to the algorithm is a wide baseline image pair and the output is a set where one element consists of eight tentative correspondences between detected regions, these are the best candidates to compute epipolar geometry between images. There are well-known techniques how to compute a fundamental matrix and reconstruct 3D coordinates. Several of them were tested to find the method that is fast and rather precise. Several ideas for future method combination are mentioned.