On the Fusion of 3-D Reconstruction Techniques

The commercial 3-D laser scanners provide high quality reconstructions of different surfaces. However if the reconstructed surfaces are made of diffusing materials and/or exhibit self occluding properties then errors are guaranteed in the output reconstruction. In this paper, we try to rectify this situation by fusing the 3-D reconstructions generated by a 3-D laser scanner and other 3-D reconstruction techniques that may not exhibit similar reconstruction errors. We propose a fusion technique that employees a 3-D alignment criterion and a local quality assessment technique as well. The fusion decision in the proposed technique is guided by the extracted contours from the input images of the object under-reconstruction. The approach can be applied to different 3-D reconstructions from different sensors. The approach exhibits promising results on the fusion of the space carving technique and the 3-D reconstruction by a 3-D laser scanner.