Modulating the Size of Backprojection Surface Patches, in Volumetric Stereo, for Increasing Reconstruction Accuracy and Robustness

This paper concerns volumetric stereo methods, which compare the back projections of the acquired images onto a hypothetical surface patch in order to reconstruct the imaged surfaces. In particular, it introduces a size-modulation of this patch so that its projection area in the acquired images is invariant to distance and rotation. It is shown, and explained why, that performing this modulation results in superior accuracy of the volumetric reconstruction than retaining the patch size constant, as it has been to date practiced. The proposed extension to the hypothetical patch operator is compatible with the existing volumetric approaches to stereo.