Dense Reconstruction by Zooming

Abstract

Reconstruction by zooming is not an unachievable task. As it has been previously demonstrated, axial stereovision technics allows to infer 3D information, but involves very small triangulation angles. Accurate calibration, data matching and reconstruction have to be performed to obtain satisfactory modelling results. In this paper, a new approach is proposed to realize dense reconstruction using a static camera equipped with a zoom lens. The proposed algorithm described in the following sections is divided in three major steps: First of all, the matching problem is solved using a correlation algorithm that explicitely takes into account the zooming effect through the images set. An intensity-based multiscale algorithm is applied to the feature points in the first image, to obtain unique point correspondences in all the other images. Then, using pixels matched by the previous method, an iterative process is proposed to obtain a sub-pixel matching. Finally, the 3D surface is reconstructed using image point correspondances. The modelling algorithm does not require any explicit calibration model and the computations involved are straightforward. This approach uses several images of accurate regular grids placed on a micrometric table, as a calibration process [1]. Complete experiments on real data are provided and show that it is possible to compute 3D dense information from a zooming image set.