Vizualizace objemovych dat a stinovani.

Volume data rendering consists of projecting the 3D information of a model into a 2D image and of shading. Depending on which illumination model is being used, shading is calculated from the global scene parameters and from the local properties of the model, which are estimated by the gradient and the normal vector. The most common method, the grey-level gradient method using central difference of the neighbouring voxel intensity, generates in some cases artefacts which are caused by wrong orientation of the normal vector and appear as dark areas in the final picture. Another method uses the binary information about the voxel classification to determine the gradient. In this way, a correctly oriented gradient is obtained, but the final picture contains staircase shaped artefacts caused by a limited number of possible normal directions. The thesis analyses situations where the above problems occur. Two new methods for gradient estimation are suggested, the reversed gradient method and the angle difference method, wihich can even be combined with each other. The reversed gradient method is based on orienting the normal vector towards the light source so that the whole surface is illuminated, which prevents dark artefacts caused by the wrong orientation of the normal vector. The angle difference method estimates the gradient by combining the values calculated by a binary method from the classification of voxels in the 26-neighbourhood and the values obtained by a different method, e.g. the grey-level gradient method. The angle between the normal vectors obtained by both methods is calculated. If the angle is greater than a suitably chosen limit angle then the gradient obtained from the classificaion is used. Otherwise, the gradient determined by the other method is used.Available from STL Prague, CZ / NTK - National Technical LibrarySIGLECZCzech Republi