The warping step has produced two warped volumes and from the source and target volumes and . Any practical warp is likely to misalign some features of and , possibly because these were not specifically delineated by feature elements. Even if perfectly aligned, matching features may have different opacities. These areas of the morph, collectively called mismatches, will have to be smoothly faded in/out in the rendered sequence, in order to maintain the illusion of a smooth transformation. This is the goal of blending.
We have two alternatives for performing this blending step. It may either be done by cross-dissolving images rendered from and , which we call 2.5D morphing, or by cross-dissolving the volumes themselves, and rendering the result, i.e. a full 3D morph. The 2.5D approach produces smooth image sequences and provides the view and lighting independence of 3D morphing discussed in section 1.1; however, some disadvantages of 2D morphing are reintroduced, such as incorrect lighting and occlusions. Consequently, 2.5D morphs do not look as realistic as 3D morphs. For example, the ``missing link'' of figure 5f lacks distinct teeth, and the base of the skull appears unrealistically transparent.
For this reason, we decided to investigate full 3D morphing, whereby we blend the warped volumes by interpolating their voxel values. The interpolation weight is a function that varies over time, where ``time'' is the normalized frame number. We have the option of using either a linear or non-linear .