Abstract: | A common problem in the literature of range data analysis is to segment a smoothed discrete surface by estimating the sign of the mean and gaussian curvatures at each point. Smoothing, however, introduces a distortion which can corrupt the curvature signs near the boundaries. In this paper we model gaussian smoothing by a diffusion equation, a general mathematical framework particularly useful for a scale-space analysis, and introduce an adaptive shape-preserving boundary condition for diffusion smoothing of range images. We show how this condition is more general than similar techniques found in the literature.
|