MIP Results

Figure 4.9: Local Maximum Intensity Projection for non-sequential samples a) new voxel in front of z[] b) new voxel between z[] and zb[]

Figure 4.10: MIP variations of the vessels in a hand: MIP, Depth-shaded MIP, LMIP.

The MIP algorithms discussed above have been implemented as a Java applet (http://www.cg.tuwien.ac.at/research/vis/vismed/RT-MIP/). The rendering times in table 4.2 have been measured on a PII/450 PC using Sun's Java virtual machine of JDK 1.1.6 for Windows with a just-in-time compiler. The first value gives the time for rendering data which was preprocessed independent of the viewpoint, the second value is the time for rendering direction-dependent preprocessed data. As timings for LMIP depend on the chosen threshold, measurements for the worst case are given. A C++ version of the MIP rendering algorithm has shown an approximately 30% better performance than the Java version.

Figure 4.10 is provided for a visual comparison of the discussed variations of MIP. It shows a 256x256x100 angiography data set of a hand rendered with MIP, DMIP, and LMIP using the presented voxel extraction approach.

Table 4.2: Frame rendering times (with direction-independent/direction-dependent preprocessing.

data set	MIP	MIP windowed	depth-shaded	LMIP
mr_angio 256$^2\times$64	169ms	27ms	209ms	$<$305ms
	94ms	21ms	118ms	180ms
mr01 256$^2\times$74	163ms	42ms	224ms	$<$288ms
	106ms	32ms	146ms	198ms
mr03 256$^2\times$124	254ms	13ms	338ms	$<$455ms
	219ms	11ms	267ms	370ms

mailto:mroz@cg.tuwien.ac.at.