PAIRWISE COMPARISON TECHNIQUE - A SIMPLE SOLUTION FOR DEPTH RECONSTRUCTION

A new technique dramatically simplifies the analysis of matching and d epth reconstruction by extracting three-dimensional rigid depth interp retation from pairwise comparisons of weak perspective projections. Th is method provides a simple linear criterion for testing the correctne ss of correspondence for a pair of images; the method also provides a description of a one-parameter family of interpretations for each pair of images that satisfies this criterion. We show that if at least thr ee projections of a volumetric object are known, then a three-dimensio nal (3D) rigid interpretation can be inferred from pairwise comparison s between any one of these images and other images in the set. The 3D interpretation is derived from the intersection of corresponding one-p arameter families. The method provides a common computational basis fo r different processes of depth perception, for example, depth-from-ste reo and depth-from-motion. Thus, a single mechanism for these processe s in the human visual system would be sufficient. The proposed method does not require information about relative positions of eye(s) or cam era(s) for different projections, but this information can be easily i ncorporated. The method can be applied for pairwise comparison within a single image. If any nontrivial correspondence is found, then severa l views of the same object are present in the same image. This happens , for example, in views of volumetrically symmetric objects. Symmetry facilitates depth reconstruction; if an object possesses two or more s ymmetries, its depth can be reconstructed from a single image.