Abstract: A method and an apparatus reduce the temporary random access memory required when determining disparity values for at least two stereoscopically recorded images with known epipolar geometry, in which a disparity is determined for each pixel of an image. Path-dependent dissimilarity costs are calculated on the basis of a disparity-dependent cost function, and compared, in two runs for a number of paths which open in the pixel. The disparity-dependent cost function evaluates a pixel-based dissimilarity measure between the pixel and the corresponding pixel, according to the respective disparity, in a second image. The path-dependent dissimilarity costs for a first predetermined set of disparities are calculated in a first run for a number of first paths and in a second run for a number of remaining paths, and the corresponding path-dependent dissimilarity costs of the first paths and of the remaining paths are accumulated for a second predetermined set of disparities.

Abstract: A method and an apparatus reduce the temporary random access memory required when determining disparity values for at least two stereoscopically recorded images with known epipolar geometry, in which a disparity is determined for each pixel of an image. Path-dependent dissimilarity costs are calculated on the basis of a disparity-dependent cost function, and compared, in two runs for a number of paths which open in the pixel. The disparity-dependent cost function evaluates a pixel-based dissimilarity measure between the pixel and the corresponding pixel, according to the respective disparity, in a second image. The path-dependent dissimilarity costs for a first predetermined set of disparities are calculated in a first run for a number of first paths and in a second run for a number of remaining paths, and the corresponding path-dependent dissimilarity costs of the first paths and of the remaining paths are accumulated for a second predetermined set of disparities.

Abstract: In an apparatus and method for representing computer-modeled objects comprising a plurality of adjacent polygon surfaces, the image impression in the corner points of the polygon surfaces is calculated according to a local illumination model, whereas the image impression of the remaining pixels is interpolated according to the Gouraud shading method, and a color or brightness value is additionally calculated for simulating a surface structure for each pixel, the value taking into consideration the influence of the surface structure on the image impression without necessitating a complicated calculation of the local illumination model in each pixel.

Abstract: The invention concerns an image-processing method and system for the simulation of structure in relief in image of objects having at least one polygonal surface. The polygonal surface is divided up into a large number of raster elements, and for each raster element, an inclination value determining the inclination of the local surface normal is read out of relief-structure matrix. From the inclination value and either the normal to the polygonal surface or the local normal to the surface, the inclined local normal is calculated and the brightness and/or color of each raster element calculated as a function of the direction of view of the observer, the illumination conditions and the angle of the inclined local normal.