6 Results

These methods and tools were tested by creating images based on several qualitatively different data sets. The following will discuss particular computed images, in order to prove that the volume visualization methods work and to discuss the importance of some parameters.

Figure 1 shows the human CT-scan volume rendered using the Maximum Intensity Projection, MIP, compositing where as Figure 2 shows XRAY style compositing. Levoy's classification and compositing equations were uses to extract surfaces. The results require user-defined density values $\rho_v$ that appropriately represent the surface. Several images with different values of $\rho_v$ are shown, demonstrating how $\rho_v$ affects the rendering of the head of the visible human CT-scan volume. After trial-and-error with the value of $\rho_v$ two surfaces of the volume were successfully extracted : Figure 3 displays the skin surface; and Figure 4, displays the skull surface. Note that slightly different values for $\rho_v$ give different surface closures, as shown in Figure 5 and Figure 6. The effect on skin selection is that extra components are rendered of top of the recognizable human skin surface. The face looks as though it is leaking, some air bubbles seems to come out, especially around the mouth. The effect on the skull is to miss some parts of the surface : the skull looks too skinny !! These two last results are satisfactory but not optimized. However, other degenerate cases of selecting an inappropriate $\rho_v$ can give weird and amazing results. For example, in Figure 7, the density $\rho _v = 55$ seems to want to wrap a surface around the head but fails to show any concrete surface features. Another example, Figure 8 shows the rendered image using $\rho _v = 148$, which happens to be between the best densities for the skin and skull surface extractions. The result is that both surfaces are visible from a single selected density value. This image is not produced using double classification, but only rendered with a density value that is between surfaces density values.

Double classification is illustrated by Figures 9 and 10. The difference in Figure 10 is that two colours were used, a greenish one for the skin and an reddish one for the skull. The skull was made transparent by assigning a semi-opaque opacity, $\alpha_v=0.25$. These results are convincing, both surfaces are extracted successfully and the interior one can been seen through the exterior one.

Figure 1: MIP compositing

Figure 2: X-ray like compositing

Figure 3: Skin surface, $\rho _v = 140$

Figure 4: Skull surface, $\rho _v = 180$

Figure 5: Skin surface, $\rho _v = 100$

Figure 6: Skull surface, $\rho _v = 170$

Figure 7: Surface extracted by $\rho _v = 55$

Figure 8: Surface extracted by $\rho _v = 148$

Figure 9: Two surface extraction

Figure 10: Two surface extraction

Figure 11: Skull rotated by $40\deg$

Figure 12: Skull rotated by $70\deg$

Figure 13: Skin rotated by $70\deg$

Figure 14: Two surface rotated by $40\deg$

Figure 15: Tomato skin $\rho _v=75$

Figure 16: Tomato interior $\rho _v=30$

Rotation around an axis was implemented. Figure 11 and 12 show the skull at two different angles. Figure 13 shows a rotation for the skin surface, and Figure 14 shows it for both the skin and skull surfaces.

All these rendered images use the contrast enhancement, which is performed as a post-processing on all the images obtained from projection.

Some other data sets were also examined but not as fully. Figure 15 and Figure 16 shows Levoy's classification on a volume representing a tomato. Figure 14 represents the exterior skin of the tomato, where as Figure 15 represents the interior media of the tomato