Surface Reconstruction with Multiresolution Discontinuity Analysis

Abstract

The goal of surface reconstruction is to reconstruct a smooth surface while avoiding smoothing out discontinuities. In this paper, a new algorithm for surface reconstruction is proposed which can locate and identify discontinuities while reconstructing a smooth surface from a set of sparse and irregularly spaced depth measurements. This algorithm uses the wavelet transform technique to induce a multiresolution approach for recovering discontinuities. In particular, the wavelet modulus maxima representation is used which allows correlation between wavelet coefficients at different scales. These correlations can be used for feature correspondence across scales. By using this multiresolution information, the estimation of locations of discontinuities is refined. The performance of the algorithm is investigated and compared with a recently published bending moment-based algorithm. It can be seen that our approach can locate and preserve discontinuities while ensuring smoothness in most of the regions.