Variational segmentation by piecewise facet models with application to range imagery

Abstract

Unlike conventional photographic images that are characterized only by light intensity or signal energy, laser radar (ladar) range data, or other terrain imagery, contain distance information. Range images are traditionally processed for their three-dimensional content. Previous innovations in partial differential equation (PDE) and total variation based segmentation techniques show good results for conventional images. Fast, efficient variational segmentation techniques that use higher order image models are needed for the preprocessing of such data for applications including target detection and acquisition, compression, and image modeling. We develop a variational segmentation algorithm using higher-order piecewise smooth models for multichannel range imagery such as ladar images. The algorithm is computationally stable, and a fast solution may be found using the fast Cholesky decomposition and a modified binary search tree.