"... In PAGE 7: ... Furthermore, no footprint simplification was applied, and we only tested open facades. The test results shown in Table1 should therefore be regarded as just an indication of what can be expected from the algorithm. The reason for this is, that at this time, the implementation is not robust enough to handle every input model.... In PAGE 7: ... In an ideal test case, the original model is part of the first category of objects, and the facade is part of the second group. It is clear from Table1 and Fig. 6, that the facade is a much more efficient occluder geometry... ..."

"... In PAGE 4: ... 3.4 Functional Performance Table3 shows the number of clock cycles necessary to perform the RTL simulation on three different Polygon implementations. It can be noted that there is a very large communication overhead due to the asynchronous protocol and the interface serialization.... In PAGE 4: ... As explained earlier, the SDL process model may have several inputs, but, since there is only one input interface in the VHDL entity generated from SDL, the reception of these signals is serialized. Table3 also shows that when the number of processes in an SDL description increases, the overhead associated to the protocol also increases. The performance is 15 or 30 times slower with... ..."

"... In PAGE 20: ...1 Spatial Subdivision Results We first constructed the spatial subdivision data structure (cell adjacency graph) for each test model. Statistics from this phase of the experiment are shown in Table1 . Column 2 lists the num- ber of input polygons in each model, while Columns 3 and 4 contain the number of cell regions and boundary polygons, respectively, generated by the spatial subdivision algorithm.... ..."

"... In PAGE 35: ... In the case of AutoShip, we start with a very simple form model of a ship (for no particular reason, we could very well start with a rough function/behavior model). This simplest form model (G1) is composed of a set of boxes and has 106 polygons, roughly within the estimate of the initial polygon budget (See Table1 ). At this point, we can think of a simple behavior model (B1) that can accompany this simple form model, which starts to navigate from a random position and orientation and changes its velocity and angular velocity at random periods.... ..."

"... In PAGE 8: ... Assuming that all of the extracted models are composed of closed polygonal surfaces, we can compute enclosed volume as a a signed sum of the pyramids with a base composed of the i apos;th triangle and a top vertex places at the origin of the dataset. 32 Then V 1 6 N poly X i=1 A i 1 3 (v 1 i + v 2 i + v 3 i ) N i ;; (17) Table1 lists values of polygon count, surface area and total volume, for the models extracted from scalar volume datasets (V 1 and V 2 ), before and after the level set algorithm is applied to the volumes. We note that the polygon count drops, because of the simpli ed form of the nal extracted triangular mesh.... ..."

"... In PAGE 14: ...76 Table 1: Spatial subdivision statistics. Table2 shows results of the beam tracing phase. The first two columns list the test model and how many input polygons it has.... ..."

"... In PAGE 14: ...76 Table 1: Spatial subdivision statistics. Table2 shows results of the beam tracing phase. The first two columns list the test model and how many input polygons it has.... ..."