A method for making a three-dimensional surface reconstruction of an object from two or more bi-dimensional images of the object, so called 2D images, comprising: a) providing a three-dimensional dataset, so called 3D dataset, of the object; b) generating a 3D model using the dataset as described in a); c) providing motion information to build a 3D model over time; d) generating a 3D reconstruction from said at least two 2D images; and e) using information from the 3D model over time to correct the 3D reconstruction over time. A corresponding apparatus and computer program are also disclosed and claimed.