Reconstruction of Human Foot from Multiple Camera Images with 3D Active Contour models

摘要

Shape-from-silhouette (SFS) is a popular method for 3D shape estimation. It is used in many 3D reconstruction applications, the limitations of this method, however, are that the approximated shape can be very coarse if the number of cameras is small, and if the object is not convex, the reconstructed 3D model by SFS may not equal to the object. In this paper, we propose a method to improve the shape approximation by combining multiple silhouette images. The improvement is achieved by applying 3D active contour models in an intensity voxel space. The intensity voxel space is initialized by normalized correlation between multiple camera images. We also apply standard model from a database to revise the foot model, then the robustness of our method is improved. We demonstrate reconstruction of 3D model of human foot from multiple camera images with the proposed method. Then the error evaluation shows our method can generate more precise 3D model than traditional SFS algorithm.