Space carving acceleration using uncertainty measurements

Abstract

This paper presents a method for obtaining a fast but rough 3D object reconstruction. This reconstruction will contain enough information to determine a minimum complementary view set that can refine it more accurately. Thus it is possible to take advantage of the space carving algorithm simplicity. This algorithm is fast and easy to accelerate by means of hardware and software techniques, but cannot easily manage the uncertainty derived from the segmentation process. In the proposed method, uncertainty is projected onto the voxels and computed with them when new views are processed. In this way, a measure of the reconstruction certainty is obtained, identifying the regions where more information is needed in order to be resolved. © 2009 Springer Berlin Heidelberg.