Abstract

Based on the growing demand for 3D visuals in various fields, presenters have the potential to communicate more on the information. One of the promising way to produce 3D media files is by converting 2D images to 3D images where this conversion even involves missing information. Any method used for conversion should give an efficient approach based on cost and storage space of 3D media in realistic application by using depth map. These maps can be anticipated either from single 2D or from multiple 2D views. Here the work presents a MATLAB based image conversion from multiple views based on the computation of a scarce depth map. The 2D to 3D conversion is possible with the multiple views obtained from uncalibrated hand-held cameras without any knowledge of the earlier camera parameters. The result that is obtained by reconstructed depth map is consistent with ground truth depth data and few options are explored to tackle this problem and the bound of these methods are established in this paper where the input is given as the image sequence of the stationary scene from different views and gets the depth of the image using camera from which we ultimately reconstruct the 3D scene. Here Epipolar geometry is prescribed to find the essential matrix and match correspondences between various images.

Paper Details