We present a method for the reconstruction of a 3D real object from a sequence of high-definition images. We combine two different procedures: a shape from silhouette technique which provides a coarse 3D initial model followed by a multi-stereo carving technique. We propose a fast but accurate method for the estimation of the carving depth at each vertex of the 3D mesh. The quality of the final textured 3D reconstruction models allows us to validate the method.
