We present a novel method for synthesizing texture over a 3D model from 2D photographs of the modeled object. By considering an optimization approach, we subdivide the mesh into patches and assign, to each patch, a stack of possible textures taken from the input images. For each image stack, we employ a multilabel graph cut algorithm to generate a single texture. With this simple and flexible approach, it is possible to create textured models for use in computer graphics, overcoming some problems in camera registration and geometry reconstruction.
Type