3-D object modeling based on surface reconstruction and integration using optimal polygonal approximation

We address the problem of 3-D object modeling that is suitable for graphic display by deriving an optimal approximation to the true object surfaces. A new model generation method is proposed, which obtains a mesh representation of the surfaces by integrating multiple range images with an optimization algorithm based on closeness to the original image data. The method is a relaxation algorithm that achieves accurate shape descriptions gradually modifying and propagating mesh arrangements iteratively. The paper first presents the formulation within the framework of regularization by redefining the minimization energy function in which both the data closeness and surface smoothness are considered. Next, we give the detailed procedure of the method repeating the surface reconstruction using the chosen regularization parameter at each stage. The search vector field corresponding to range data is generated alternately. The method has been applied to real range images to verify the applicability, and results are reported in the paper.