No abstract

No abstract

In this article, we focus on the comparison of the passive techniques of multi-view 3D reconstruction, namely the following techniques : Passive Stereo vision, Shape from Silhouette and Space Carving. Available data in the Passive techniques are no more than one or many images taken from different point of views (using one or several cameras). These images will be used in order to render the three-dimensional scene. Our study is based on the quality of the solution (3D Model obtained) and the calculation time. In passive stereo vision, the quality of the results is evaluated in terms of the value of the re-projection error. Also we will evaluate each technique separately. In the end, to enjoy the benefits of the techniques studied we proposed a method based on the extraction of silhouette images and the matching between images to make full 3D object reconstruction. The results of our own implementation of the different techniques studied and the proposed method enable to compare and evaluate these techniques, and to show the quality of the results obtained by the proposed method.Keywords-3D reconstruction; passive stereo vision; shape from silhouette; space carving.

In this paper, we are interested in the problem of Euclidean 3D reconstruction of unknown objects by passive stereo vision method. Our method is based on the combination between Harris and Sift interest point detectors, to take advantage of the power of these two detectors, which will be useful when matching step, as a key step for 3D reconstruction, In order to have a sufficient number of matches distributed on the images. These matches will be used to estimate the 3D points (the projection matrices will be estimated after calibration using 3D Calibration Pattern). Finally, a 3D mesh is constructed by 3D Delaunay triangulation, applied to the 3D points reconstructed. Experimental results prove that this method is practical and gives satisfying results without going through the propagation step.

No abstract