For boundary voxel specification, either the iso-surface metaphor, or a predefined and explicit segmentation mask is used. In the first case, the technique described in chapter 4.3 is used. The 6-connectedness of the resulting sets of boundary voxels is useful for exploiting coherence during compression of the contours. Boundaries of objects which are defined using a segmentation mask, can be extracted in a similar way and also result in 6-connected sets of voxels.
Although best compression efficiency is achieved for surface-like voxel sets, truly volumetric objects can be extracted and compressed in the same way. This is especially useful for the visualization of spatially complex structures, like vessels in medical angiography data sets or complex chaotic attractors in the field of dynamical systems [4]. For the extracted voxels, attributes like voxel position, data value, gradient direction and magnitude, and application specific attributes are stored. When only the display of shaded surfaces is desired, storing voxel position and gradient direction is sufficient.