Abstract: Passive methods for three-dimensional reconstruction of a scene by means of image data are generally based on the determination of spatial correspondences between a number of images of the scene recorded from various directions and distances. A method and a device are disclosed which provide a high reliability in the solution of the correspondence problem in conjunction with a low computational outlay. Image areas for determining the correspondences are determined within a plurality of images forming at least two image sequences. In preferred embodiments, a parameterized function h(u, v, t) is matched to each of the image areas in a space R(uvgt) defined by pixel position (u, v), image value g and time t. The parameters of the parameterized functions are used to form a similarity measure between the image areas.

Abstract: Passive methods for three-dimensional reconstruction of a scene by means of image data are generally based on the determination of spatial correspondences between a number of images of the scene recorded from various directions and distances. A method and a device are disclosed which provide a high reliability in the solution of the correspondence problem in conjunction with a low computational outlay. Image areas for determining the correspondences are determined within a plurality of images forming at least two image sequences. In preferred embodiments, a parameterized function h(u,v,t) is matched to each of the image areas in a space R(uvgt) defined by pixel position (u, v), image value g and time t. The parameters of the parameterized functions are used to form a similarity measure between the image areas.