Contouring Results

Preprocessing statistics for several subvolumes of the Visible Human datasets are given in Table 1. Note that the size of the seed set is 1-2 orders of magnitude smaller than the total size of the data. Tree Size denotes the number of cell labels stored in a segment tree data structure for the seed set. Again, storage overhead is much less than the total size of the data.

Data	Resolution	Seed cells	% of total	Distinct Seeds	Tree Size	Preprocessing (s)
Male Head	512x512x181	2217568	4.72%	3294	8922765	596.8
Female Head	512x512x208	2276972	4.21%	3154	8548789	669.7
Male Pelvis	325x150x270	1110420	8.56%	2517	6602785	183.2

Table 1: Data and Preprocessing Characteristics

Surface extraction data are presented in Table 2. Note that the seed cell approach results in a very consistent performance rate as measured by triangles/sec, indicating that the average complexity of surface extraction is linear with respect to the size of the resulting surface. This claim is further supported by the graph in Figure 3, which plots performance for multiple isovalue queries on a subvolume of a femur of the Visible Male.

Data	Isovalue	# Tri	March Time(s)	Our Time (s)	Speedup	March Tri/s	Our Tri/s
Male Head	600.5	1329920	323.52	28.40 (s)	11.39	4111	46828
same	1224.5	2110962	322.53	47.20 (s)	7.05	6545	44723
Female Head	600.5	1415154	381.10	35.60 (s)	10.71	3713	39752
same	1224.5	2196390	385.59	56.23 (s)	6.85	5696	39060
Male Pelvis	1224.5	1879088	94.20	45.56 (s)	2.07	19947	41244

Table 2: Surface Extraction Timing Statistics

  
Figure 3: Linear performance for surface extraction from Male Femur

  

(a)	(b)	(c)

Figure 4: Isosurfaces from the Visible Male (a-b) and Visible Female (c). Speedups attained were up to 11 times (over a brute-force surface extraction approach)