Remeshing visual hull approximation by displaced butterfly subdivision surfaces

This paper proposes a new procedure for generating a displaced butterfly subdivision surface from multiple images. First, point geometry of the target object is recovered by combining LDC(Layered Depth Cube) surfel sampling scheme with the concept of visual hull based on the input images. Then the subdivision surface is generated approximating to the recovered point cloud. We use a variant displaced subdivision scheme, where scalar displacement, in the direction of a local normal, is computed via the MLS(Moving Least Squares) approximation. The resulting subdivision surface is a mesh with subdivision connectivity providing a high-quality and efficient approximation to the given images. And it is able not only to represent a natural level of detail structure of the surface, but it is also to be memory-efficient by taking advantage of smoothness properties. Experimental results show the quality of our algorithm.