Automatic extraction of deformable part models

摘要

I deseribe a system that first segments an image into parts, and then fits deformable models to range data associated with the image. Models are represented by using modal dynamics applied to volumetric primitives, which significantly improves the computational complexity of both model recovery and subsequent processing. The segmentation procedure uses two simple mechanisms: a filtering operation to produce a large set of potential object parts, followed by a quadratic integer optimization that searches among these part hypotheses to produce a maximum-likelihood estimate of the image's part structure. Once a segmentation has been produced, a volumetric description is obtained by a fitting procedure that minimizes squared error between the range measurements an the model's visible surface. For simple part shapes, it is possible to compute the deformable model's parameters using only the shape of its symmetry axes.